New issue
Advanced search Search tips

Issue 877587 link

Starred by 1 user
Status: WontFix
Owner:
enne@chromium.org
Closed: Sep 21
Cc:
m...@chromium.org
enne@chromium.org
khushals...@chromium.org
gayane@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

19.4%-43.8% regression in loading.desktop at 584985:585047

Project Member Reported by m...@chromium.org, Aug 24 
See the link to graphs below.
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Aug 24 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=877587

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=ef2a093ebbae4c9ca49098adc1cc4d9f362b3c36fa01654c7dee1719434317b3


Bot(s) for this bug's original alert(s):

Win 7 Nvidia GPU Perf
mac-10_13_laptop_high_end-perf

loading.desktop - Benchmark documentation link:
  https://bit.ly/loading-benchmarks
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Aug 24 

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/15de57ce640000

Comment 5 by m...@chromium.org, Aug 27 

Re-running with wider bisect range.
Project Member

Comment 6 by 42576172...@developer.gserviceaccount.com, Aug 27

Cc: gayane@chromium.org enne@chromium.org
Owner: enne@chromium.org
Status: Assigned (was: Untriaged)
📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/13924283640000

Set an expiry date for unused histograms (Conflicts.*, ThirdPartyModules.*) by gayane@chromium.org
https://chromium.googlesource.com/chromium/src/+/6466c6d2c02ed48eb1f4c53bce45e9e00c261e00
216 → 246.2 (+30.2)

Turn on OOP Raster on mac bots by enne@chromium.org
https://chromium.googlesource.com/chromium/src/+/ee35f329c8e615505326a3b4df9ae4b7764037ae
238.5 → 254.2 (+15.67)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Comment 7 by enne@chromium.org, Aug 28

Cc: khushals...@chromium.org
Components: Internals>Compositing>OOP-Raster

Comment 9 by enne@chromium.org, Aug 28 

The Windows regression looks unrelated to my "mac bot" change.  Bisecting to see what moved that.
Project Member

Comment 10 by 42576172...@developer.gserviceaccount.com, Aug 29 

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/110173bd640000
Project Member

Comment 12 by 42576172...@developer.gserviceaccount.com, Aug 29 

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/119e5f6d640000

Comment 13 by enne@chromium.org, Sep 21 

This looks strongly like shader compilation and like similar things we saw on Android.  This regresses cold starts but not warm, and so it seems like the shader cache issues described here: https://docs.google.com/document/d/1aAmFiBlrv2wV8QE1ySdHoNi6sPspWyrJwnolk8QnfNU/edit#heading=h.fqka2siwy8yx

Sent a trace pinpoint job to verify: https://pinpoint-dot-chromeperf.appspot.com/job/1644b7a8e40000

Comment 14 by enne@chromium.org, Sep 21

Status: WontFix (was: Assigned)
Yes, sure looks like shader generation to me.
►

Sign in to add a comment