All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=910250

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=7b141ea7fbac853550b85bd40496cb80c03d269c184c5e21a6138e138b54c115


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

linux-perf

rendering.desktop - Benchmark documentation link:
  https://bit.ly/rendering-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/102e63c8140000

/cc @sullivan: The ref/actual metrics are behaving unexpectedly (ref metric is improved in the same time period where actual metric has regressed). Have you any ideas why this may happen?

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/102e63c8140000

I've seen this once or twice before, but never gotten to the bottom of it. My first guess would be that there is a change to the metric that gives different results in older chrome vs ToT, but pinpoint should bisect into metric changes, and it couldn't reproduce.

Adding Ehsan and Sadrul, any ideas what could be happening here?

I will look into this tomorrow.

So, the fact that ref and tot changed in different directions is not super surprising for tasks_per_frame metrics.

For example, if ref and tot have different process architectures (e.g. in-browser viz vs separate viz process) a change can introduce new IO tasks in just one of them. So, almost always I verify tasks_per_frame changes with cpu_time_per_frame.

In this case, cpu_time_per_frame changes in the same direction for ref and tot:

https://chromeperf.appspot.com/report?sid=3f1a8be975eb98ee18daee80e175eb90e169af79026e4b566da0bab3cb55c290&start_rev=607977&end_rev=613074

However, I cannot explain the change itself. I do not see any telemetry/metric changes in the CL range...

Can someone from infra take a look, please?

1- What did change bot ref and tot graphs? There is no metric change in the CL range. It doesn't look like a change in bots since the graph has a land/revert pattern...

2- Also, if relevant, why pinpoint couldn't find the reason?

Thanks!