New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 596615 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner: ----
Closed: Apr 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 3
Type: Bug



Sign in to add a comment

telemetry tests for image decode performance

Project Member Reported by jzern@chromium.org, Mar 21 2016

Issue description

In a follow-up to a change I made to improve webp decode performance in Chrome on Android Egor pointed out that it might be useful to have a regression test for this if possible [1]. I don't have any familiarity with this system, so I was curious what the current state of things was.

Salient points from the bug:

> Are these benchmarks on the perf dashboard so that regressions are automatically analysed?
>

No. I don't believe decode performance for images is being captured currently. We are tracking webp performance during development and at release, but it's true that won't catch a regression due to e.g., build changes.

...

Feel free to create a bug for continuous regression testing if you need it. telemetry@chromium.org group would have some ideas how to make it in a maximally blessed way. I guess that would be like one of our raster benchmarks, but there is always a risk of seeing deprecated ways to write a benchmark...

[1] https://bugs.chromium.org/p/chromium/issues/detail?id=416951#c9
 

Comment 1 by pasko@chromium.org, Mar 23 2016

Cc: sullivan@chromium.org nednguyen@chromium.org
+Ned, +Annie for telemetry and tracking perf regressions. Even though Telemetry may be overkill for a microbenchmark like this. Maybe it could be similar to cc_perftests, but a quick glance to the perf dashboard suggests no data for more than a year.

Lost in the dashboard.

I would suggest to ask telemetry@chromium.org

Comment 3 by jzern@chromium.org, Apr 4 2016

Status: WontFix (was: Untriaged)
Thanks for the links. I can see an improvement in commits in the range of the change I was concerned about (a1/n9/s5) and that it caught another issue later (#593670), which is nice.
I'll close this as it doesn't appear there's any work to be done.

Sign in to add a comment