All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=867978

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=254912c02ae6ea946f69c6ba38ce2c7c1022d850a50580e4f8882595b1ebf003


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

mac-10_13_laptop_high_end-perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14ad18d3a40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14ad18d3a40000

Free more resources in gpu/oop raster when idle by enne@chromium.org
https://chromium.googlesource.com/chromium/src/+/da08e1808cb0835dd647407869b84e6f30a745d0
97.44 → 127.7 (+30.26)

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

On mac, this patch at r577956 reverted ericrk's at r572391.  What's curious is that frame times on this test appear to be entirely unchanged after ericrk's patch and changed after my revert.  The patch revert is reverting two oneliners, so I feel pretty confident that it's doing the right thing.

regressions from clearing fewer gpu raster resources (non-android):
https://chromeperf.appspot.com/group_report?rev=572391

regressions from clearing fewer oop raster resources (android):
https://chromeperf.appspot.com/group_report?rev=576180

regressions from reverting both:
https://chromeperf.appspot.com/group_report?rev=577956

There's only one test that appears to have been affected negatively by this patch.

The original bug at the beginning of all of this was trying to fix a 5% input latency regression.  I think the memory savings are worth that original regression and this should just be closed WontFix.