All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=794973

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=a848013a5d22f787585ea350bff51a47028f7695718cc9b118e15076991b4408


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

chromium-rel-mac-retina
chromium-rel-mac11-pro

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1487b2da040000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/1487b2da040000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1484f6d4840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a644f4840000

Kicking off a few more bisects, hopefully they'll get around the error.

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/14a644f4840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14cef85d840000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14cef85d840000

converted vertex shaders to device coords by ethannicholas@google.com
https://skia.googlesource.com/skia/+/e7e81c15c144b8133f696d0744ed9f7e8d06e936

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

ethannicholas, can you take a look? Looking at the results in #10, it seems pretty clear that your CL regresses performance on this benchmark.

I just dug into this again and continue to be unable to replicate the slowdown. Both Skia benchmarks and Chrome's smoothness.gpu_rasterization.tough_path_rendering_cases are unchanged with and without my patch on both systems I tested it on. There's no obvious reason why my patch would affect performance and the other GPU team eyeballs I've had look at it agreed that it should not affect performance.

Given that there were a lot of issues with the bisectors identifying the wrong CLs the day this bug was filed, is there any chance something else is going on?

 Issue 794972  has been merged into this issue.