New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 643211 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Closed: Sep 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

3.7%-31.2% regression in smoothness.top_25_smooth at 415249:415307

Project Member Reported by mustaq@chromium.org, Sep 1 2016

Issue description

See the link to graphs below.
 
Cc: skyos...@chromium.org
Owner: skyos...@chromium.org

=== Auto-CCing suspected CL author skyostil@chromium.org ===

Hi skyostil@chromium.org, the bisect results pointed to your CL below as possibly
causing a regression. Please have a look at this info and see whether
your CL be related.


===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : scheduler: Throttle timers in out-of-view frames
Author  : skyostil
Commit description:
  
Throttle timer tasks to 1 Hz in out-of-view or invisible frames. This
reduces the power consumption and performance impact of these tasks,
which are not likely having any visible effect on the page anyway.

BUG= 616519 

Review-Url: https://codereview.chromium.org/2023033003
Cr-Commit-Position: refs/heads/master@{#415278}
Commit  : 7fec349e59e40dd9beee5f7eafa9e0db5c4ac1bd
Date    : Tue Aug 30 12:47:29 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev   N  Good?
chromium@415260  10.7789  0.339584  8  good
chromium@415272  11.0195  0.86224   8  good
chromium@415275  10.9666  1.10597   5  good
chromium@415277  10.4762  0.321776  5  good
chromium@415278  13.0311  0.247018  8  bad    <--
chromium@415284  12.818   0.139047  5  bad

Bisect job ran on: mac_retina_perf_bisect
Bug ID: 643211

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests smoothness.top_25_smooth
Test Metric: mean_input_event_latency/Docs  (1 open document tab)
Relative Change: 17.77%
Score: 98.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_retina_perf_bisect/builds/1630
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9002709183993968928


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6633758856314880

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!
Components: Blink>Scheduling
Started a trace for the smoothness.top_25_smooth regression: https://codereview.chromium.org/2305943003

(I think we can ignore the sync scroll regression since that test is pretty artificial. The fact that it regressed there too might help to debug this other one however)
Looks like the tracing button isn't working because this is a TBMv1 benchmark :\

I did some local testing and was unable to reproduce the regression neither on linux or mac.

Here are a couple of traces from the build bot:

Without throttling: https://00e9e64bacf94f881d866bf8d8178d7204901904ee9841964f-apidata.googleusercontent.com/download/storage/v1_internal/b/chrome-telemetry-output/o/trace-file-id_5-2016-08-29_02-34-21-97181.html?qk=AD5uMEsdgd857QxohGjoYzO2y4gn0WGXKoDmE0Q5_pEy3yzv28k1mRK0VzrAlE53hlXChX_bDM2-YV2r_1m15U4_P-UXeFDerbAqC7cyT6Ro6H6ON5sTjMZbEt3ZYqm8juPbA8odBuRdNQRDnQkus5ZQh_TVNWqKUf5k2-iA_1eaViMxcoJEPO3vsy3AIb38JeVJH61SXJTYTkXbTNoXgTe6COviKQfA5uO_lHw2uoYyUAlz9wgShrRXgrBC0uVvIRa6NOK5wl9TVa89S0422eLb9wo8BbmwGATCPku6Bml5oDhjHbYd3gdwd1IxeJpyjBZ3Ows6mTfL9tzdA5hc8JjJYbMnRHZIBA64YCU7XZ_QGcj0KTbzG-s6_ken6eAhzDxGx9cUBSodEywz3NgqOy6vWxfRYOuKhpCriHzUJ14zbL6q9kXeBEno9imOE5FoaFPPRyHHYeirB7iL7k9sr3kQ9hPKhJBGFcBugn9vcW9VL07tOYMDQdiGF4vG4b5R1CPoAg23Lhx0kxQ_udwDigtVyknU3y7g_hehpiJ7_Odx0lBwFVR6yDGxxVdH-XHPzivxCY58idSR4tCrZdgLW-ZG2vriigzqvcviCyMoQWPQo0QSfC8RCepF9HblVOF252nsoxe7deYq3ElHIfdcNXTsgCTnz3lQ3TgVrV-MoHt5Q0sYGvTd2-NCNr-sgmfDv2xVp6MjjWAHjBuNlBvBe4hOK7Mp_KbDvUg7T7Le9Q2SRvLsRuAWYPEZHY1z9uJypJv2umoUkveGpeO1_iXIVYEGxBDqj39k7NOJvkgiwoSbl7WAySUswyNbCU9g3UIgk3aubT2FIUgNcB1Ak-TcgBIytPxcj59kHr3Iommz0lkJn52kt2vUOeA

With throttling: https://00e9e64bacda9bd4b7e6e52bbd2cc10395c91a4eea070ba272-apidata.googleusercontent.com/download/storage/v1_internal/b/chrome-telemetry-output/o/trace-file-id_5-2016-09-05_13-09-05-1805.html?qk=AD5uMEsi1n1RNZkHaiDvPFMzKE3Fo_GWO4SbHhYM3qb8e18HZlg9awuSTyE9NEq_Yx680bSiBcX2XIX5AJw5Kr7s9IwkQkMFThh9tlz7WbjDp8qEkE5k-6kF311O1_uI9jdayynCbnCb8RMfZS7r7dRP34oG6umlO9LexeZ3dnEAO6xS-zESXEk3JVOx4JAF5Z2JTzhmaG8w1k31psAMhhGwkZ6d8WbVI4zUQlt5V5dhOzHZ5XFX07O-2Fb_5oFfWRxjybh8vpHt6DldjR9WoTCIVEGHcE8C7ob1mTeQxLNQkMi19UEt-QnAlDwLSqKIwjyTjU8GxVWeNkwXF493cw_WLlLlBMo196oSAM4eeTTU3uVzId9jP6tMAWD5f-6rlC0JfpIJf6GGPmahGjURgdMGWpgkVcNj75yxais6v-ZtZQcXRBMKYTpopDKXFURidumSYEKXYGK1vUqFF92Z46jp7hPVJmOa68kPIRxRbmRB1A7FU7gA4FmWG6jGzN5qMpLT9bd_CU5upDAf7uwVLHvWdto0wv9O_j7vT6IjN5JSxWJHZAW_OzrTlawLXTkRBgMcI-poJrjuM-WZD6LKYJSDq1p5x4csCPzeNLRnha4gNOXLIzfNuF-52MA34KX1szl_Iy-J8Tv_cpJCVJ0XBWwSz8-Ycu7T6bjO-8IH4sUmhOu9hkNsmY_oVYYIu2ypuryaO8R-lP04ezS4D1s6JKPtZvIXXvSKREwybVRHg3fGfum3mj4-0icyQIFCSz1BfIVPcrk_J3bHnKDP5inT0IDxYQnyrY3prOtFf0t4GRFgSm4gP1cB4XOnfx7pqSs8E99wmL2Ze62YSdBx__PB2QddFH3xzzK3XdVxY834gc8TF_xjAwbN4zk

I think the difference comes down to the benchmark starting a little sooner when throttling is enabled, and in doing so it manages to overlap with a long main thread task (see the attached screenshots). The actual scrolling phase is equally smooth in both traces, so I think the regression is due to the measurement starting earlier.
without_throttling.png
9.9 KB View Download
with_throttling.png
9.5 KB View Download
Cc: hpayer@chromium.org u...@chromium.org
Status: WontFix (was: Assigned)
Here are some memory traces:

Before: https://00e9e64bace091dea55a80a022ab8a15b6e2301a1edc45e8ab-apidata.googleusercontent.com/download/storage/v1_internal/b/chrome-telemetry-output/o/trace-file-id_3-2016-08-30_09-18-34-81696.html?qk=AD5uMEtlClhPSjdhyHafwc6ZS4l0X6Q8gqSuxrrdsY9eMRDO1t3YDIY5OzBnXHkOmiZdyL0P0ckh_IjEmp7VKp6yb26tD-DlX00lv_liP6iO4GN3BGzzY5Q4rW9iVl34iXXXC222b5bCGddxO_6ySwrEGNe90kIbscgjfSWp8_q46gJrqQ6hKa1VXZ1x_iFJ7omF8yZNTBK04EsH5EZErpRt-1hHXpUNlOfdevoZvZ6M0douLoeAX_KmqlLTgTKHv-HBCmC8Ml6TyhdFpHFn8NgQWl3LVSK79_hvh8n9q8Ryk86y0K8YGcE6pK9BmF3H0316KYQecmxZZ1iXSR4Rh0KpxaH8I7n8yOweKhOcws0waLnKpEki1J8KuJkYSNvNCDIK2cL2eua_s4gVXu8ClqIiaQ7NDNpu-K1R-ntZsFS_pLSJfKoJWNOMogFVIuBAe-N5iAHysaegSGGSoG5K3IwcHqpvRfLZKRFmgHGFOMqruxfsW0sfvQI7A_N9lzDeM091Rl1QQU1zRepsbW6M2cm7qHkwub5fP-Nc_wrA6swf0hKyrX45lGqJICMNA5b6Im0ZsNIXMrTygNNiTcHvavD9aVx34hDsGN_0glmGV7o-Ie83QSFJ4DiGSeLo05m7PMKXaXh-__aM96iRbrw7UU1c73jRPPozCTtCzN-rW-0hzolDP5wiJ9xA9GoI5hLO6h3NG225INyx9Dcj6Y3MdyM0KB3wsE2Jy3iaABs7nq-ZdyeMIeM4JrLvA3K93oVX6XoFwVJ2yOkDOl6hNElmFluO4vXgwNqliiA3dGO74qETQwXqcSWNFDb1zsdJFvwQ5g7OP4oKi1mxF6FZyvhvjyuanHvlEyx-J__xWOSc9ziOsRV7fDdqeSXD54joRTjH1_Vois-ylj7mosO3GzvcwJHqy54_7_1HGg

After: https://00e9e64bac35177dfbf65b4316eff04b26ab1bc6bfc0cab3e9-apidata.googleusercontent.com/download/storage/v1_internal/b/chrome-telemetry-output/o/trace-file-id_3-2016-08-30_14-00-54-42088.html?qk=AD5uMEsWJOYqFBEdfBhko0HHyI-s8t1EWvb4l05SDyljT8Z9-k22U1m-0MTfeLY36tysZ-9aTHnXkixl1p1WsOO8f_n6TybPZ7j7uBqyr5cF7zDViR2RhY0Xa5pUWYux_HcyvMcmOJPjkwIZourjtSMIenyHl3aBqxfEn96sPpWSiri2wtkqhaNT_JLXSEeudIZJqCXq3PQd9QOfleTS8tqOsy93GkDYC8joDADgd0r4V_35wo9rC0UfvYOA8NHdNc4esEiZgXxePVLsUn_LCLOlFzdkU4UkX63Q6SHxIKMnuQTpLPIvjnfM9hJYqUrtySS1DfjeHlrz110XF7NN-e4Z2QKZOfI1YuNQ04kbB95Bu7hHVIAwnpE6d1aQMuasGuEUwWpWu2y4mKuZe3HGc34o5GChxvlpmcjN9N426C8mw1FEW8Kus6xfr8uymZVESlNpS3Ch5qqzyueu1kyjjSe4Egtu-o-56V-kVrjO6BimeOIHToKox_Mnhlv0tiDNddTi5yBmmNwGsDkmMHOgaznM5NP_9OoK99FjjI3ALWItbQyvh7nC-A_mRVQ2RBrTBID8WOLk7OkuXvCMkoJmzz3DkWVrTu3zVAwbgZ-LB7onp1qmfHoC-8Y2K-DHZkd3KlJ1rrbEsbOVZj-FBrLs1ePBITchKwbIGTIM4o5lhtXvawTXr_75AL4kc4eALo2ifp3jNsPG37PNCU2eggqVp7YmnKZwvu0So0mkyskXaNJO4qz5cMgpCTmSXTKPrBne6BceFT8TZ7qaVj5XzBzJi-xl54xpi2fw-SIbWwnVulhO-8sKb6y0pIwtfWljEIZIvOpo8jJ2TgFDkzNaKSzvl0cdbaB8f44IXOSjWUVo31XPtELvQ39xoeHrw2tpl3DaYx2SuJOEGjiAJL7AZgcWLHtgJGAKAiN3qw

The overall v8 heap size drops by about 2 MB but the old space becomes about 100k bigger (which triggered this alert). Seems like a good trade-off to me -- but +cc some experts to make sure this is legit. WontFix in the meanwhile.
FYI, bisect bot just merged in 645548 which has more memory regressions.
Ah, thanks. Looks like on gmail v8-gc-scavenger_avg went up. Anyone know if that's good or bad?-)

Comment 9 by hpayer@chromium.org, Sep 12 2016

Hmm, I think you are just having bad luck here. You can ignore the scavenger regression.

Sign in to add a comment