All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=855407

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=abea530dac852eb83508abc7ed78f74e1ba3ab13ad2ff5c89c025c3715d27bdc


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

android-nexus5

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14818697240000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14818697240000

Set TouchStartExpectedSoon only after we seen a blocking event by chaopeng@chromium.org
https://chromium.googlesource.com/chromium/src/+/4a2ebc8dad89ca0042d0babd619f4ea35e3ac072
1.68e+06 → 4.196e+06 (+2.517e+06)

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

Hi Sami, do you have idea how this happens?

I've started a detailed trace recording on pinpoint: https://pinpoint-dot-chromeperf.appspot.com/job/1482b23f240000

Let's take a look at it tomorrow.

The NextAction date has arrived: 2018-06-26

The job died, rerunning it...

The NextAction date has arrived: 2018-06-27

The pinpoint fails consistently.

Here are the traces from before and after the regression:

https://00e9e64bac8758eabaa425050668da65a544e90d05b3efabb8-apidata.googleusercontent.com/download/storage/v1/b/chrome-telemetry-output/o/background_search_google_2018-06-19_17-56-08_26152.html?qk=AD5uMEseD8OAf2sPq4crtSnnlqmfc6iTY4JXVZ8y_zTVMiIipKDm3lphHl5-WPyvLZXh6HEPDiernpwi3cwnTeCpibYRCuUj6BzyF-mxhRm9NHQyu793IvlJwNFBWRhx0CUfVbVi2vhPP-GKQspS9Fq5GTytLj7tKSpiXlt5xq8LB6dn-aFNrisYPa0W3CJ4EA8OI1BDnR8r4WOfTWz6ILAYhnromRc7o5VYfAsF_n9IhhFNZJzJzbSiFLwe4_VS53qZU_pAvWwOoDGPfw2ZGt3VJn6uZNqlb7A38DXaGNvtG31-BWlbqvRNeoDOw6G0Swl-RaL3PRgeQnF44fLDVn6Bwccv6ix09zD_smLUUNp2Inay7JYwTmIWoyTY-OQWJi1GOZA0i2rzJ9xJ8Qv6uRgzSPR5GSVQNqCXEL1_YLZB8BurfXURwRMl_EcamrYz028p-tX9BJP0LSi3K3AzL6wce00oByMoDTRCmLvc3kaTCR4Laai0ofEGI1l37O3YQ0f_JEUk9mVo2IN8_Rvlt-mJM41ao5ezhG1E5GLkzVH7phoJYJkR9Aen_GWQWhaFNUtnQqIW_NSeeco9y1TpS0ySzBOBfeZv_C8KZfkkhTxN_GeZ4L3k-l6OnLzBsgbBl1z5o6WMUxRmf9MYHGPWQ-T5gbvm2EbGnuvaIHwTMGd6TwwVfVb0Y5Rbitwu3E-ted_YNCSrNFi0NU-LYFj3Xl-e_wWDX-QmwDSA8Szz1iS1zT_eaHu_0pk3ylxRqfEpIHmjfwculILxI_OpfCrlkCrvfgSAwfiIcsh17--w7pu4_0zwW2OZ5cM


https://00e9e64bac6305ed877d2888f27eaa65865cc068eaddc9c888-apidata.googleusercontent.com/download/storage/v1/b/chrome-telemetry-output/o/background_search_google_2018-06-20_04-48-19_29315.html?qk=AD5uMEudwsbZuwpSAFxYJlXwaRCFH8cNxlPsmcFOh6A-yZukCkIxh4O35ngK3cZstkCq2dBNVJb-g6fVUwG5l629QSncw2D9jHL7Y1x2t2gEERcy44MnlrS_srBTRv_lNz6znWsbb9ry8mD3sAEEQaSTvGjfsYB6R_JF56XruzBsyT26peUsZgRItUuCrHSDTRc3eaTy9HpylTOeHxhU79pljgBmFx3aZP918ZwwsbZrb_bhEs6-hBAIcFb1iIH9nvYzNVRdnnwqIrZjsacdPVhJKBpUYHkbFXsVX4hBosYwCkcRO4C0ee_Bxf0NqK-oLJlFD_XNw8v0QHDPhsW8Yrwxu_kXIK2XwkCQGlrTZTHlWGQdY7auScTFSeLMDPs11V9pU6G8ro59Nan3EmNoCj2kU3QRnYVgbQl34ogtxhmp_czlytFg88pHhZe2Zt6oN_kfl4V4dCGvlA3qbKb6RRFy6h25faPhi_Y1r0fHZbprp8AR2yzr8_6yCmpeSE-3uT-OPbvijUHA2aG4BskD7zYzSz69QkDRKEc-ZtgNrYaaVnOCM6NzkclFnfck7Xqc7ukOOE4hnofYvd_h04TaSC3wYF7ydGQHuY3h8Fd-bkChmOQqcfF22XkVxG5iJkEHxx4EGApzIv-3cCGbEyEDS68jfS6rItKfKPNqgk4jOCaglao6iciBkVRdVhpypFJjENX4hbbWQ5MeKSpQqcgZuVRDAUmEi5I-QAmbihyTGQL_2SByKHuYIhqFRloPmDbHFdYxDUy_8w0nefJliHsdlrE_eshev0yUCNo1FjYB7LuYVwUzSB3BkUQ

In the latter case looks like the GPU process has allocated 3 extra buffers (2x1MB + 1x2MB) for the renderer. My hunch is that the scheduling changes made the page end up in a slightly different state than before, which causes these buffers to get allocated sooner. I doubt this is an actual regression.