tree 400d0f98c0f46b7053cb01cfc060596b46ff76c8
parent 6f2e304938da568b473d43dbe09c8abc147c7541
author Findit <findit-for-me@appspot.gserviceaccount.com> 1619154328 +0000
committer Copybara-Service <copybara-worker@google.com> 1619155377 -0700

Revert "Add SubframeShutdownDelay experiment"

This reverts commit 3447a7a52eecea83be3ddb03ae65d4c2f8df5b43.

Reason for revert:

Note: It is reported that sheriffs cannot submit CL created by Findit
(crbug.com/1187426). A workaround in the mean time is to abandon this
CL and create another revert CL.

Findit (https://goo.gl/kROfz5) identified CL at revision 875466 as the
culprit for failures in the build cycles as shown on:
https://analysis.chromium.org/waterfall/culprit?key=ag9zfmZpbmRpdC1mb3ItbWVyRAsSDVdmU3VzcGVjdGVkQ0wiMWNocm9taXVtLzM0NDdhN2E1MmVlY2VhODNiZTNkZGIwM2FlNjVkNGMyZjhkZjViNDMM

Sample Failed Build: https://ci.chromium.org/b/8849198831280035280

Sample Failed Step: content_unittests

Original change's description:
> Add SubframeShutdownDelay experiment
>
> This change adds experiment SubframeShutdownDelay, which delays
> shutdown of subframe renderer processes by a few seconds. This
> experiment aims to increase the chance for subframe processes to be
> reused, particularly when navigating between pages that share a common
> subframe domain. Successful subframe-process reuse saves the cost of
> process recreation, which can take ~3 seconds of CPU time per
> navigation. This savings needs to be balanced against the memory cost
> of keeping each subframe renderer alive a few seconds longer.
>
> To this end, the experiment has multiple branches with different
> delays, in order to determine the best-performing method:
> * constant: a flat 2s delay
> * constant-long: a flat 8s delay
> * history-based: a 0-8s delay; the median
>               TimeSinceReusableProcessDestroyed in the past 5 subframe
>               navigations, with "no reuse possible" counting as 0s
> * history-based-long: a 0-8s delay; the 75th-percentile
>               TimeSinceReusableProcessDestroyed in the past 5 subframe
>               navigations, with "no reuse possible" counting as 0s
> * memory-based: a 2s delay when available memory is under 8 GB, and a
>                 2-8s delay when it is higher, up to 16 GB
>
> These will be enabled via Finch, and can be manually enabled by
> appending command-line flag:
>    --enable-features="SubframeShutdownDelay:type/constant
> where "constant" can be any of the above experiment labels.
>
> Design doc:
> https://docs.google.com/document/d/1x_h4Gg4ForILEj8A4rMBX6d84uHWyQ9RSXmGVqMlBTk/edit
>
> Bug: 894253
> Change-Id: I64c64532d6185412d7f3f97bf9917de9574aecc4
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2431587
> Commit-Queue: Jesse McKenna <jessemckenna@google.com>
> Reviewed-by: Alex Moshchuk <alexmos@chromium.org>
> Reviewed-by: Olivier Li <olivierli@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#875466}

Change-Id: Ieab89a6fe5aba60abdf08f0b9b86e8e160c694e1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: 894253
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2847642
Reviewed-by: My Nguyen <myy@chromium.org>
Commit-Queue: My Nguyen <myy@chromium.org>
Owners-Override: My Nguyen <myy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#875526}
GitOrigin-RevId: 6a9440ddbb65844905eb7f442623b2431df20d07