All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=917259

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=dfab7aa5cb4ff40a7c0badfa7db30e97b9394e6530414ba2c76bdf4ecebd6d11


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

mac-10_13_laptop_high_end-perf

blink_perf.bindings - Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14b3d926140000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/14b3d926140000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14fc5bf1940000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/149b765e940000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/149b765e940000

Add SequenceManager to field trial testing config by carlscab@google.com
https://chromium.googlesource.com/chromium/src/+/84ecbf7c0b1eca668dba1471a50f065c6d558737
post-message: 295.3 → 366 (+70.68)

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

Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/14fc5bf1940000

This is strange the regression only seems to be on one mac bot, although I note one of the windows bots got more noisy around the same time.  Anyway I instrumented chrome's task posting and ran this test and it seems to do a lot of cross thread immediate task posting.  It's not obvious why that would be worse one the high end mac laptop.

https://chromeperf.appspot.com/report?sid=3d0f4253a54d17811ff5de630a4d60eb9de3b6452769e7c7574a27476585e77c

Hmm I was misreading the graph, the regression is on the low end bot too.

 Issue 920372  has been merged into this issue.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/163c4d2e540000

📍 Found significant differences after each of 3 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/163c4d2e540000

[scheduler] Remove expensive task blocking logic. by altimin@chromium.org
https://chromium.googlesource.com/chromium/src/+/071306963377eb133a94b80a6bbcce949f99e7b8
post-message: 291.4 → 265.1 (-26.29)

Adding feature flag for 'accessibility language detection'. by chrishall@chromium.org
https://chromium.googlesource.com/chromium/src/+/4a4225f68380d640a87837dd4fa11827192bb4ef
post-message: 275.6 → 270.7 (-4.807)

CacheStorage: Hold a reference to CacheStorage from Cache by bashi@chromium.org
https://chromium.googlesource.com/chromium/src/+/7bc86423ba2bdd56b15a5cf44089043142326dda
post-message: 267 → 275.8 (+8.824)

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

Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

The mac regression seems to be resolved now by the removal of the expensive task blocking logic (although the bot[1] hasn't run the test since Friday for some reason). I don't have any theories as to why that patch would have had this impact on performance -- ideas welcome :) The regression seems to have been mac-only too [2].

[1] https://chromeperf.appspot.com/group_report?bug_id=917259
[2] https://chromeperf.appspot.com/report?sid=9d2787fac49c8efd202af8ecff80d934976c5c3795dd5d0af5237ce2185abbb9&start_rev=620208&end_rev=624732

That's unexpected.  The task cost estimators might account for at least some of the improvement.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/17ddd21a540000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/17ddd21a540000