Issue metadata
Sign in to add a comment
|
4.1%-9.5% regression in page_cycler.tough_layout_cases at 379578:379607 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Mar 9 2016
=== Auto-CCing suspected CL author eae@chromium.org === Hi eae@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 : Pass Windows font rendering preferences from browser to blink Author : eae Commit description: Change the way Blink queries Windows font settings to work reliably with the sandbox. Instead of calling SystemParametersInfo directly from Blink in FontPlatformDataWin do so in the browser process and pass the desired flags to Blink using the WebFontRendering interface. Also, use the value should_antialias_text in Chrome's RendererPreferences object rather than checking the SPI_GETFONTSMOOTHINGTYPE Windows system parameter directly. Similarly use the subpixel_rendering field in RendererPreferences rather than checking FE_FONTSMOOTHINGCLEARTYPE to control use of subpixel text. In addition to ensure that these values are queried correctly this helps guarantee that Chrome and Blink use the same setting for text rendering. TBR=jam@chromium.org BUG= 419003 Review URL: https://codereview.chromium.org/1763803002 Cr-Commit-Position: refs/heads/master@{#379589} Commit : e3fe85330f1fbad6cb02cd0c5ceb09498e50f349 Date : Mon Mar 07 18:22:20 2016 ===== TESTED REVISIONS ===== Revision Mean Value Std. Dev. Num Values Good? chromium@379577 734.593875 8.080694 5 good chromium@379585 735.705 8.187963 5 good chromium@379587 735.969563 5.497814 5 good chromium@379588 734.888313 11.328744 5 good chromium@379589 808.154188 0.910886 5 bad chromium@379593 802.402938 1.642786 5 bad chromium@379607 804.360563 2.585986 5 bad Bisect job ran on: win_8_perf_bisect Bug ID: 593102 Test Command: python src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --also-run-disabled-tests page_cycler.intl_ar_fa_he Test Metric: warm_times/page_load_time Relative Change: 9.50% Score: 99.9 Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/1871 Job details: https://chromeperf.appspot.com/buildbucket_job_status/9018730612700624720 | 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 label Cr-Tests-AutoBisect. Thank you!
,
Mar 9 2016
eae@, there appears to have been a regression for page load time on windows 8 esp. for non-Latin-script languages (Arabic, Hewbrew, Korean, Thai, Devanagari, Chinese, ...) This regression is much clearer on some sites than others: http://hindi.oneindia.in http://www.mydiyclub.com What's special about non-Latin fonts on Windows 8?
,
Mar 9 2016
Nothing really. I'm guessing that before we fell back on the default settings on the windows 8 bots as we didn't have permission to read the preferences. We didn't change any of the logic, only the way we read the user settings from within the sandbox. Certain types of AA is more expensive than others. As long as this is limited to a subset of tests for a single bot I'm not to worries. If we see this on either the win7 or win10 bots though please let me know and I'll investigate.
,
Mar 9 2016
===== BISECT JOB RESULTS ===== Status: completed ===== SUSPECTED CL(s) ===== Subject : Pass Windows font rendering preferences from browser to blink Author : eae Commit description: Change the way Blink queries Windows font settings to work reliably with the sandbox. Instead of calling SystemParametersInfo directly from Blink in FontPlatformDataWin do so in the browser process and pass the desired flags to Blink using the WebFontRendering interface. Also, use the value should_antialias_text in Chrome's RendererPreferences object rather than checking the SPI_GETFONTSMOOTHINGTYPE Windows system parameter directly. Similarly use the subpixel_rendering field in RendererPreferences rather than checking FE_FONTSMOOTHINGCLEARTYPE to control use of subpixel text. In addition to ensure that these values are queried correctly this helps guarantee that Chrome and Blink use the same setting for text rendering. TBR=jam@chromium.org BUG= 419003 Review URL: https://codereview.chromium.org/1763803002 Cr-Commit-Position: refs/heads/master@{#379589} Commit : e3fe85330f1fbad6cb02cd0c5ceb09498e50f349 Date : Mon Mar 07 18:22:20 2016 ===== TESTED REVISIONS ===== Revision Mean Value Std. Dev. Num Values Good? chromium@379577 1067.737 14.417007 5 good chromium@379585 1065.2465 15.330138 5 good chromium@379587 1080.062 19.04069 5 good chromium@379588 1076.4325 51.899487 5 good chromium@379589 1266.4735 16.570675 5 bad chromium@379593 1264.5215 12.837942 5 bad chromium@379607 1258.6865 14.782322 5 bad Bisect job ran on: win_8_perf_bisect Bug ID: 593102 Test Command: python src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --also-run-disabled-tests page_cycler.intl_hi_ru Test Metric: warm_times/http___hindi.oneindia.in_ Relative Change: 17.88% Score: 99.9 Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/1873 Job details: https://chromeperf.appspot.com/buildbucket_job_status/9018648224360667056 | 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 label Cr-Tests-AutoBisect. Thank you!
,
Mar 9 2016
===== BISECT JOB RESULTS ===== Status: completed ===== SUSPECTED CL(s) ===== Subject : Pass Windows font rendering preferences from browser to blink Author : eae Commit description: Change the way Blink queries Windows font settings to work reliably with the sandbox. Instead of calling SystemParametersInfo directly from Blink in FontPlatformDataWin do so in the browser process and pass the desired flags to Blink using the WebFontRendering interface. Also, use the value should_antialias_text in Chrome's RendererPreferences object rather than checking the SPI_GETFONTSMOOTHINGTYPE Windows system parameter directly. Similarly use the subpixel_rendering field in RendererPreferences rather than checking FE_FONTSMOOTHINGCLEARTYPE to control use of subpixel text. In addition to ensure that these values are queried correctly this helps guarantee that Chrome and Blink use the same setting for text rendering. TBR=jam@chromium.org BUG= 419003 Review URL: https://codereview.chromium.org/1763803002 Cr-Commit-Position: refs/heads/master@{#379589} Commit : e3fe85330f1fbad6cb02cd0c5ceb09498e50f349 Date : Mon Mar 07 18:22:20 2016 ===== TESTED REVISIONS ===== Revision Mean Value Std. Dev. Num Values Good? chromium@379577 4703.842 152.925303 5 good chromium@379585 4609.43975 14.503894 5 good chromium@379587 4627.8 19.878396 5 good chromium@379588 4592.01 30.971815 5 good chromium@379589 5330.519 22.927887 5 bad chromium@379593 5326.02275 17.717097 5 bad chromium@379607 5310.227 75.454717 5 bad Bisect job ran on: win_8_perf_bisect Bug ID: 593102 Test Command: python src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --also-run-disabled-tests page_cycler.tough_layout_cases Test Metric: cold_times/http___mydiyclub.com Relative Change: 12.89% Score: 99.9 Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/1874 Job details: https://chromeperf.appspot.com/buildbucket_job_status/9018648191595033376 | 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 label Cr-Tests-AutoBisect. Thank you! |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by qyears...@chromium.org
, Mar 8 2016