RootLayerScrolling increases frame times of scroll on https://mail.yandex.ru by ~60%
Reported by
lo...@yandex-team.ru,
Apr 16 2018
|
|||||||||
Issue descriptionUserAgent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/65.0.3325.181 YaBrowser/18.4.1.320 (beta) Yowser/2.5 Safari/537.36 Example URL: https://mail.yandex.ru Steps to reproduce the problem: 1) Record WPR on https://mail.yandex.ru 2) Run smoothness benchmark on this WPR ( gpu-rasterization is off ) Check with flag --enable-features=RootLayerScrolling and with flag --disable-features=RootLayerScrolling What is the expected behavior? Frame times are equal with and without RootLayerScrolling. What went wrong? Frame times increase by ~60%. Does it occur on multiple sites: N/A Is it a problem with a plugin? No Did this work before? N/A Does this work in other browsers? Yes Chrome version: 65.0.3325.181 Channel: stable OS Version: 6.1 (Windows 7, Windows Server 2008 R2) Flash Version: Shockwave Flash 23.0 r0 We have 25 web-sites in our smoothness top 25, but we observe perfomance drop only on https://mail.yandex.ru. https://00e9e64bac2d7433b2975540b119cd7cfdc1a15981b5f8b75d-apidata.googleusercontent.com/download/storage/v1/b/monorail-prod.appspot.com/o/16%2Fattachments%2F66fc1004-76c4-45af-acdf-597acd85fb31?qk=AD5uMEsI6GCUsRZf3IIdC7GyOznv82R2NQzdAQcsgCkBcOHCUv-l2i5AMi5zJ47oC55LTjTWYyulHJOKA5nvNVc7wegJnutJJeM6Ub4b9iEuCjqgaa-PxnjFsEHRi4E9nfH-ipoXUN8Ubbzw8IRltD66BExpWZUH_hkxrixdCABtjMGZRcffkqQdX1zk1Ai-OEQX-VJSRE1WvErAtkX9lIaCPqPahx4fzIQqGYFYtzt1vW5QcVxKBLsYjbZuDcMVAli9sIOvRPGMyZJFQq3crU5T_-VtkrQn7dqZpCuM7mvRV0UAYzLX_bkcMV15uKXGIpjk_0ghLzPcCKY1ISRGauo4l7JHLSJLokYfY4rSYPGa4FiJ6q_wX8Uko8K01kQ_EtSEGwEFXI1mgFr__3tzDO-J3D7fZEYT_1vrVl5JGBcFxKlrOx15LWPzvkpi3QAAKhayQ3FlziMhm5yLNT6nBzapsY1msjKmwIQf-Rebj6MMrJht0iu9y47TfCsJ2X_9bLnjLWicO9dPhWGbclG6MVd8c28qd2HOMwd3wXewbgA_Fd8EjZgf90zcmipY6YPhjqtayl0tfIsF27Um_1kjuculM3BLuG62pUZqNgKXXMDM0QgmlrnnuQcOimJBWdjN0dmRkD9vbSZ4_naXZu8gptiLJyRLkN3Q5CvAKG3U9xgb1SaSIekqInLfJLj9FvEDAWhAdI2Gg0qQBjdW2mIAJ7ewyLYXMyMg2u0Akc88bL_Xb4lvqHAYCmps0vV1mHeDl8wTqdj2GPiYi06I0l9WaOSdOJagxPFRRg https://00e9e64bac3290b697e3c1e66677b42d06db9c1ff1bce948c9-apidata.googleusercontent.com/download/storage/v1/b/monorail-prod.appspot.com/o/16%2Fattachments%2F702c8755-ac97-4f1f-b6b6-b1a06c47c76c?qk=AD5uMEtrCAi2dTguvLk6QWNp3sA8PjhBydqcxsNH2Z0MYTJnng8QpfrTSwOVw2uUerK-FTHfbnSRiQ-H1KRLobZbqGzDTjm9s1EW9ZyWe1rAAPTt74MEiOaQzAx58RiWJti_ucLCuzWMk59CBXuPEpNvPTFykuMPy5cDhJadkBNeKhJA3h92PL6VNlsXU8DufYo4iBQDMNAhJqtsbMKjQyj7-ZdDoISDvmu1vLymDbjaLhcXqieZPMQBcvQPEMJAk_ki-Wut2K34KSPh9N_-jlDoKfP756hBpLOW--ie8NRe6g80lgZPOskj131tBVjXHggIFjokuUT9GUb-LFVZaeI1kGpfCFJge8Pkb6klZTzVgD0s6DbRiEp2X8Inilkl1j2fA-CvZYcYVDT6j-vUYBVD1E0jVV5ALfC8NPo3lHjMRrOPRL_kzvdJ68wmzn85A1LVqwus6YbjchI0-BPBjkhCiqz_5QqpDIc660fLZOGqPwm2hxcQZMY5ssOdanMjF25EZHx0fFXu2i-GG-hE3x_aK-_NFr0WTnqPUqmH4wzzF4660cyt3EKOFEQqxBf7Fm-F_tvJ7SKl_KVP-mpAYsfYXXlAq3Y1buTfos-zUtfx5DVxFi0ad0zOGw3AnpfiiLceIlK6lMaPn48L4ibtlQXzuXT6F0sISjTrMKDKJbsKA9ZKos3BU_gGSDwAlEHA6W34ADBIuQTMHPziBFXHjPiDcvlBblWnjPbBguFdtb24SKL8tIvE5SJNZkESgfJT0ioi2PFtHoE_4GKQGvtR4CyZJQY5Sc6pTg
,
Apr 16 2018
Does this require logging into the site? If so, do you have a set of test credentials we can use to reproduce?
,
Apr 16 2018
,
Apr 17 2018
Yes, we have a test account with a bunch of e-mails in inbox, which allows for extended scrolling durations and more thorough frame stats. szager, can I send you that account credentials via e-mail?
,
Apr 17 2018
Yes, please send test credentials to szager@chromium.org, thanks.
,
Apr 18 2018
Adding label Needs-Feedback as per comment#4 and #5 and requesting reporter to share the test credentials to the mentioned mail ID(szager@chromium.org). Thanks!
,
Apr 26 2018
,
Apr 26 2018
I haven't been able to reproduce the perf difference. The links you provided in the original bug report appear to have expired. Could you provide more details on how you got your results?
I'm using test credentials for mail.yandex.ru, and this patch to the chromium source:
$ git diff
diff --git a/tools/perf/page_sets/top_25_smooth.py b/tools/perf/page_sets/top_25_smooth.py
index 21e24a20f70e..0e50e0295cc3 100644
--- a/tools/perf/page_sets/top_25_smooth.py
+++ b/tools/perf/page_sets/top_25_smooth.py
@@ -142,6 +142,7 @@ _PAGE_URLS = [
('http://sports.yahoo.com/', 'yahoo_sports'),
# Why: top tech blog
('http://techcrunch.com', 'techcrunch'),
+ ('https://mail.yandex.ru', 'yandex'),
]
I then run the perf benchmark with this command:
$ src/tools/perf/run_benchmark run smoothness.top_25_smooth --profile-dir=/path/to/chrome/profile/signed/in/with/test/credentials --story-filter=yandex --use-live-sites
... and then I run a second time, adding this flag:
--extra-browser-args=--disable-features=RootLayerScrolling
I have attached screen shots of the results page for both runs; the numbers for frame time are almost identical.
,
Apr 26 2018
Fixed attachment.
,
Apr 30 2018
Hello! Do you disable gpu-rasterization? Try to add --disable-gpu-rasterization flag. I attached our perf results in html.
,
Apr 30 2018
Thank you for providing more feedback. Adding the requester to the cc list. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
May 2 2018
FYI, I am looking into this, but still have not been able to reproduce the regression. My latest results (attached), with --disable-gpu-rasterization, do show a performance regression, but much smaller than what you've observed. I have been running on Linux; tomorrow I'll try on Windows. I think it's unlikely to make a difference, but we'll see.
,
May 2 2018
Our slow perf-bots are Dell Latitude E4300 notebooks. It is rather slow machine with Core 2 Duo 2.5Ghz, Intel GMA X4500, 4GB RAM and HDD.
,
May 9 2018
As per comment #8, the issue is already been taken care by szager@ and the issue need to be tested using Python 2.6, Port 80 and Port 53 UDP and TCP to support DNS traffic which is out of TE-scope. Hence, adding label TE-NeedsTriageHelp for further investigation from dev team. Thanks...!!
,
May 9 2018
,
May 10 2018
,
Aug 10
|
|||||||||
►
Sign in to add a comment |
|||||||||
Comment 1 by szager@chromium.org
, Apr 16 2018Owner: szager@chromium.org