Issue metadata
Sign in to add a comment
|
13.4%-13.8% regression in blink_perf.shadow_dom at 594754:594839 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Oct 2
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/13c1f6e4e40000
,
Oct 3
📍 Found significant differences after each of 2 commits. https://pinpoint-dot-chromeperf.appspot.com/job/13c1f6e4e40000 Roll clang 342523:343189. by hans@chromium.org https://chromium.googlesource.com/chromium/src/+/7c58d162bce2af84b2ec900fe9d66ae17c3ae8ad v0-small-distribution-with-layout: 815.1 → 794.6 (-20.52) Add UMA and UKM for forced layouts by schenney@chromium.org https://chromium.googlesource.com/chromium/src/+/fe7f8c3cc9889e4a1fd18742691d621982b2c171 v0-small-distribution-with-layout: 800.6 → 915.4 (+114.9) Understanding performance regressions: http://g.co/ChromePerformanceRegressions Benchmark documentation link: https://bit.ly/blink-perf-benchmarks
,
Oct 5
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/b942c118f1ba5aab7e3cfd3b962aa277cd851346 commit b942c118f1ba5aab7e3cfd3b962aa277cd851346 Author: Stephen Chenney <schenney@chromium.org> Date: Fri Oct 05 14:44:11 2018 Record MainFrame ratio UMA only every 30 seconds Mobile performance regressed when UMA and UKM metrics for the Rendering Core team were added to BeginMainFrame. We hypothesize that the primary cost is recording UMA on every frame for the new metrics, so this patch switches to recording UMA only every 30 seconds, or on frame destruction, when the UKM metrics are recorded. The plan is to monitor performance graphs to determine if this has the desired effect. For the ForcedLayout metric we may need to send the primary UMA metric only once per frame, not once per forced layout as is done now. That is a follow on patch. R=vmpstr@chromium.org BUG=890871,891003,891341 Change-Id: I2f4ec8bb9301cece98aef52463d671e4bebcd91e Reviewed-on: https://chromium-review.googlesource.com/c/1260443 Reviewed-by: Steven Holte <holte@chromium.org> Reviewed-by: vmpstr <vmpstr@chromium.org> Commit-Queue: Stephen Chenney <schenney@chromium.org> Cr-Commit-Position: refs/heads/master@{#597113} [modify] https://crrev.com/b942c118f1ba5aab7e3cfd3b962aa277cd851346/third_party/blink/renderer/core/frame/local_frame_ukm_aggregator.cc [modify] https://crrev.com/b942c118f1ba5aab7e3cfd3b962aa277cd851346/third_party/blink/renderer/core/frame/local_frame_ukm_aggregator.h [modify] https://crrev.com/b942c118f1ba5aab7e3cfd3b962aa277cd851346/tools/metrics/histograms/histograms.xml
,
Oct 15
Still not recovered.
,
Oct 17
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/2ecdfc5d281dfca76778a1218aca5215aaf91793 commit 2ecdfc5d281dfca76778a1218aca5215aaf91793 Author: Stephen Chenney <schenney@chromium.org> Date: Wed Oct 17 19:27:32 2018 Revert "Record MainFrame ratio UMA only every 30 seconds" This reverts commit b942c118f1ba5aab7e3cfd3b962aa277cd851346. Reason for revert: The patch caused all of our ratio metrics to disappear on the UMA dashboard. The metrics are being sent, but I suspect the sample count is too low for reporting. Also, it didn't fix the perf regression. Original change's description: > Record MainFrame ratio UMA only every 30 seconds > > Mobile performance regressed when UMA and UKM metrics for the > Rendering Core team were added to BeginMainFrame. We hypothesize > that the primary cost is recording UMA on every frame for the > new metrics, so this patch switches to recording UMA only every 30 > seconds, or on frame destruction, when the UKM metrics are recorded. > > The plan is to monitor performance graphs to determine if this has the > desired effect. > > For the ForcedLayout metric we may need to send the primary UMA metric > only once per frame, not once per forced layout as is done now. That is > a follow on patch. > > R=vmpstr@chromium.org > BUG=890871,891003,891341 > > Change-Id: I2f4ec8bb9301cece98aef52463d671e4bebcd91e > Reviewed-on: https://chromium-review.googlesource.com/c/1260443 > Reviewed-by: Steven Holte <holte@chromium.org> > Reviewed-by: vmpstr <vmpstr@chromium.org> > Commit-Queue: Stephen Chenney <schenney@chromium.org> > Cr-Commit-Position: refs/heads/master@{#597113} TBR=vmpstr@chromium.org,rkaplow@chromium.org,holte@chromium.org,schenney@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 890871, 891003, 891341 Change-Id: I4668b91a1cc7080d65ba722a2168652cd5b89416 Reviewed-on: https://chromium-review.googlesource.com/c/1286735 Commit-Queue: Stephen Chenney <schenney@chromium.org> Reviewed-by: Stephen Chenney <schenney@chromium.org> Cr-Commit-Position: refs/heads/master@{#600516} [modify] https://crrev.com/2ecdfc5d281dfca76778a1218aca5215aaf91793/third_party/blink/renderer/core/frame/local_frame_ukm_aggregator.cc [modify] https://crrev.com/2ecdfc5d281dfca76778a1218aca5215aaf91793/third_party/blink/renderer/core/frame/local_frame_ukm_aggregator.h [modify] https://crrev.com/2ecdfc5d281dfca76778a1218aca5215aaf91793/tools/metrics/histograms/histograms.xml |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by 42576172...@developer.gserviceaccount.com
, Oct 2