All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=801126

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=dccf5d1ea1c9e254fe36acf3a5dc4a1d028946aa3bea440f6a65c400082ad179


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

chromium-rel-win7-x64-dual

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14944df7040000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/119a94c2840000

Re-kicking bisect on a wider range.

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/119a94c2840000

--> khusalsagar@ any chance crrev.com/528065 is related?

Doesn't look like it. The patch above is r528065 which is outside the regression range r528100 - r528197. The only canvas related change in that range is https://chromium.googlesource.com/chromium/src/+/2f16d12eef0d92ca5768f3aed42eb923c25c63b9.

xlai@, mind taking a look in case this is related?

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11fdf917840000

I have a very hard time trying to identify this cause of this regression.

Among the two CLs I've had in the regression log, 
https://chromium-review.googlesource.com/c/chromium/src/+/857811 is removing dead code of an enum type.
https://chromium-review.googlesource.com/c/chromium/src/+/855040 is making changes to OffscreenCanvas. I look at the code over and over again and there is no modification to HTMLCanvasElement or anything that it is using is touched by that CL.

Local bisecting is not working well. I am unable to reproduce the difference in
performance.

junov@: You also have a CL that adds CHECK to StaticBitmapImage. Mind take a look at this as well?

You can remove the CHECK since it was for an investigation that is no longer relevant. But it seem unlikely that that would be the cause.

The multiple pinpoint jobs were unable to repro the regression, and the regression is not visible on any other bots.  Smells like the cause could be external factors (not the source code). Don't hesitate to "WontFix" this issue after you've exhausted all reasonable options.

The regression happened only on the Dual core bot, so the CL that caused this may not even be canvas related.  Anything that increases CPU core contention could be to blame.  

Also, I take back what I said about external factors: the reference build's performance is flat, which disproved that hypothesis.

Based on #10, #11, I really do not think this is a canvas-related issue. Among the 3 canvas CLs, one is removing dead code, one is adding CHECK, one is modifying functions that are purely used by an
unshipped feature OffscreenCanvas, not HTMLCanvasElement.

In fact, the LetItSnow performance test is a complex package that uses both HTMLCanvasElement, HTMLImageElement, SVG image and CSS. Anything that alters the CPU core or rendering can affect it.

The following CLs look like possible candidates too:
https://chromium-review.googlesource.com/c/chromium/src/+/809811 touching on Window machine
https://chromium-review.googlesource.com/c/chromium/src/+/854976 touching on picture layer
https://chromium-review.googlesource.com/c/chromium/src/+/851712 touching on CSS
https://chromium-review.googlesource.com/c/chromium/src/+/857056 touching on SVG

Assigning to the SVG CL owner. futhark@: Could you take a look at the code to see if this
is related?

I just added a couple of null checks. Highly unlikely that's causing a perf regression.

alexis.menard@intel.com: Any chance that this is related to a change in Windows? as this regression only happens in Windows.

https://chromium.googlesource.com/chromium/src/+/f3d67d226996b866053a711f59cdd465b363a63a

My change is related to tablet mode and MQ, nothing related to canvas.

The regression is not caused by canvas-related changes. Putting it back to untriaged to request help from Telemetry team to identify the real owner, possibly by a commit-by-commit bisecting on a windows dual core machine. Currently the auto-bisect bot is not doing well.

📍 Found a significant difference after 1 commit.
https://chromeperf.appspot.com/job/11fdf917840000

Remove SK_SUPPORT_LEGACY_2PT_CONICAL by liyuqian@google.com
https://chromium.googlesource.com/chromium/src/+/c030ab4a4be3e740a244055ed5a70f0d19a32033

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

I can't open https://chromeperf.appspot.com/job/11fdf917840000 so I checked https://pinpoint-dot-chromeperf.appspot.com/job/11fdf917840000

From its figure, I don't see a big jump in my CL. I only see a missing line segment that should connect the two dots (one CL before mine, and my CL). Am I interpreting the figure correctly?

I see the measurements come down to normal level in recent revs. Shall we close this bug?

Re #20, sounds good to me. :)