All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=788779

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=8b4f168836827dacf62f2478e122b30140e81f4d3fdd861c6a6241c853f0cbac


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

cros-cyan

According to ihf@, graphics_SanAngeles is a very old app and we shouldn't look at it for performance difference.  pwang@, can you remove this alert?

I think we should still get alerted, but discount alerts a lot. It is basically a very weak signal, but having it can help find out where a problem came from.

I am more on not deleting the alert completely, but tweaking the alerts so that we don't get too much noise.

Checked my email, it seems that the SanAngeles is quite stable (not much alerts generated). So I think it is fine to keep the current setting.

The fps drop in this bug seems to be a real thing?

I bisected and the commit that caused the issue is:

717e7539124dc459276385a02847b06ea1989973 is the first bad commit
commit 717e7539124dc459276385a02847b06ea1989973
Author: Kenneth Graunke <kenneth@whitecape.org>
Date:   Fri Sep 8 15:00:14 2017 -0700

i965: Use a WC map and memcpy for the batch instead of pwrite.

We'd like to eliminate the malloc'd shadow copy eventually, but there
are still unresolved performance problems.  In the meantime, let's at 
least get rid of pwrite.

On Apollolake, improves Synmark OglBatch6 performance by:
1.53581% +/- 0.269589% (n=108).

Reviewed-by: Matt Turner <mattst88@gmail.com>
    Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

It's just weird we have a FPS of 260-320, when in reality it should between 0-60 fps.  I guess it measures something that is correlated with frame rate?  In any case, let's ignore this regression for now since we are unsure about the usefulness of the benchmark.  If we find other regressions associated with that commit, maybe we'll do something...