All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=741797

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=bb0838e8eebd058fbadfcbcfe882e74053648aaccd51902089ccaae358954de7


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

chromium-rel-win8-dual

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8974249360955215776

=== Auto-CCing suspected CL author vmpstr@chromium.org ===

Hi vmpstr@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Vladimir Levin
  Commit : 1dc1c0e1ced508a0545d59cc55dbc4d16d7a8cd3
  Date   : Tue Jul 11 02:38:14 2017
  Subject: cc: Don't use high filter quality for software images.

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : system_health.memory_desktop
  Metric       : memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size_avg/browse_media/browse_media_tumblr
  Change       : 23.33% | 531234626.833 -> 407318263.833

Revision             Result                     N
chromium@485416      531234627 +- 8927773       6      good
chromium@485472      529260795 +- 7993992       6      good
chromium@485486      532532282 +- 11536508      6      good
chromium@485493      532494995 +- 11887437      6      good
chromium@485497      535097433 +- 8093932       6      good
chromium@485498      536527016 +- 9347635       6      good
chromium@485499      406263555 +- 6373071       6      bad       <--
chromium@485500      406718114 +- 8489597       6      bad
chromium@485528      407318264 +- 4753238       6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=browse.media.tumblr system_health.memory_desktop

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974249360955215776


For feedback, file a bug with component Speed>Bisection

\o/

Should we try the Android one too?

Yes please do! That would be super mysterious.

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8974227441705773248

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8974227141129591632

Here's the list of all perf effects of the Vlad's patch: https://chromeperf.appspot.com/group_report?rev=485499

I looked at doing a perf bisect of the Android changes in that range but I don't see any that don't have bugs and are not reverts of previous patches (e.g., https://chromeperf.appspot.com/report?sid=34cd808b268d0c45bf60fc0fe8298a2abeff483defb92aa46b98d89b13995d8d&rev=485518). I'm not sure Vlad's patch affected Android after all.

I kicked off a bisect of this high-dpi progression because it could be related:
https://chromeperf.appspot.com/report?sid=53291c1729e64067871803bc16c5ec3f891584f8263be90faecc2e72010732a3&rev=485530
"ChromiumPerf/win-high-dpi/system_health.memory_desktop / memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg / load_tools / load_tools_docs"

I also kicked off a bisect of this high-dpi regression:
https://chromeperf.appspot.com/report?sid=09bef2e2821eee33e4515d2651558650df480733e9834e83c960df04c48f634b&rev=485530
"ChromiumPerf/win-high-dpi/system_health.memory_desktop / memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size_avg / load_games / load_games_bubbles"

=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Christopher Cameron
  Commit : 77e966647f6b495fe44772086c709528c711fc6e
  Date   : Mon Jul 10 17:45:28 2017
  Subject: Make SkImage_Lazy::onMakeColorSpace return a SkImage_Lazy

Bisect Details
  Configuration: winx64_high_dpi_perf_bisect
  Benchmark    : system_health.memory_desktop
  Metric       : memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size_avg/load_games/load_games_bubbles
  Change       : 36.53% | 373896522.167 -> 510481998.667

Revision                             Result                     N
chromium@485402                      373896522 +- 18386785      6      good
chromium@485466                      370192614 +- 12998585      6      good
chromium@485482                      386063777 +- 44604134      6      good
chromium@485490                      380464255 +- 31523372      6      good
chromium@485491                      371838921 +- 6047441       6      good
chromium@485491,skia@9f772a44fe      377873426 +- 27873656      6      good
chromium@485491,skia@62ae664e74      384580892 +- 29402808      6      good
chromium@485491,skia@77e966647f      509309872 +- 34769605      6      bad       <--
chromium@485491,skia@d8ddf1bbb4      528195151 +- 32335096      6      bad
chromium@485492                      515207476 +- 38040262      6      bad
chromium@485494                      505320569 +- 8346454       6      bad
chromium@485498                      524156720 +- 33606690      6      bad
chromium@485530                      510481999 +- 37788088      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=load.games.bubbles system_health.memory_desktop

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974227141129591632


For feedback, file a bug with component Speed>Bisection

=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Christopher Cameron
  Commit : 77e966647f6b495fe44772086c709528c711fc6e
  Date   : Mon Jul 10 17:45:28 2017
  Subject: Make SkImage_Lazy::onMakeColorSpace return a SkImage_Lazy

Bisect Details
  Configuration: winx64_high_dpi_perf_bisect
  Benchmark    : system_health.memory_desktop
  Metric       : memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg/load_tools/load_tools_docs
  Change       : 23.69% | 29371803.1667 -> 22412944.6667

Revision                             Result                   N
chromium@485402                      29371803 +- 3445191      6      good
chromium@485466                      29478008 +- 1662159      6      good
chromium@485482                      30217552 +- 4203362      6      good
chromium@485490                      29621656 +- 3955716      6      good
chromium@485491                      29445923 +- 5036963      6      good
chromium@485491,skia@9f772a44fe      29465128 +- 5212151      6      good
chromium@485491,skia@62ae664e74      29318303 +- 2144304      6      good
chromium@485491,skia@77e966647f      22254021 +- 395698       6      bad       <--
chromium@485491,skia@d8ddf1bbb4      22310401 +- 820844       6      bad
chromium@485492                      21977201 +- 834678       6      bad
chromium@485494                      22183287 +- 449194       6      bad
chromium@485498                      22033014 +- 1131250      6      bad
chromium@485530                      22412945 +- 446304       6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=load.tools.docs system_health.memory_desktop

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974227441705773248


For feedback, file a bug with component Speed>Bisection

So I'm reading this correctly, from #3 the highq patch improve desktop memory. From 10 and 11, a skia patch about color spaces regresses bubbles but improves load tools. Did I interpret that correctly? 

If so, I think we can assume that the android improvement is likely due to the color space as well.

Yes, that sounds correct and plausible. I'm putting something together that fixes bubbles.