Android M52 - Smoothness Regression

11 views
Skip to first unread message

M, Sarathchandra

unread,
Aug 17, 2016, 2:32:48 PM8/17/16
to tele...@chromium.org

Hi,

 

This is Sarath from Amazon. We noticed that M52 Chrome has a smoothness regression compared to M51 Chrome when measured on multiple Amazon Devices.

 

Smoothness regressed by ~20% with Chrome M52. Quick search in chrome M52 bugs didn’t popup any bugs related to smoothness.

 

Could you please share the bug if you have any or suggest the process to provide this regression as a feedback to Chrome.

 

Thanks,

Sarath | TPM, AWS Silk

Ned

unread,
Aug 17, 2016, 2:34:32 PM8/17/16
to M, Sarathchandra, tele...@chromium.org, vmi...@google.com, tdre...@google.com

--
You received this message because you are subscribed to the Google Groups "telemetry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to telemetry+...@chromium.org.

Victor Miura

unread,
Aug 17, 2016, 5:55:52 PM8/17/16
to Ned, M, Sarathchandra, tele...@chromium.org, tdre...@google.com
Hi Sarath,

We aren't tracking a smoothness regression in M52 Android as far as I'm aware.  Could you please share more details?  You may file a new issue on bugs.chromium.org and assign it to myself (vmi...@chromium.org) and I'll follow up.

Thanks,
Victor

M, Sarathchandra

unread,
Aug 18, 2016, 2:22:12 AM8/18/16
to Victor Miura, Ned, tele...@chromium.org, tdre...@google.com

Thanks for the update Victor. I will file a bug tomorrow with all the information we have.

Tim Dresser

unread,
Aug 18, 2016, 9:55:10 AM8/18/16
to M, Sarathchandra, Victor Miura, Ned, tele...@chromium.org
In the bug, can you clarify what you mean by smoothness?
Which measurements in particular?

Does the regression persist in M53 and M54?

In M52, on many devices, we see a regression in the mean input event latency, but a significant improvement in the 95th percentile, which is a tradeoff we're willing to make. It's possible that this is what you're noticing.

Tim

Victor Miura

unread,
Aug 18, 2016, 12:04:22 PM8/18/16
to Tim Dresser, M, Sarathchandra, Ned, tele...@chromium.org
I may have found an issue.  Following shows a large jank regression just before M52 branch point.  This seems to be affecting only tablets (Tab UI).  Sarath could you confirm if you are seeing a regression on tablets rather than phones?


M, Sarathchandra

unread,
Aug 18, 2016, 2:02:41 PM8/18/16
to Victor Miura, Tim Dresser, Ned, tele...@chromium.org

Hi Victor,


Yes, Regression is observed on Tablet devices (We run tests only on tablets). It is visible if you take two nexus7 devices running android 5.1.1 and do a side by side comparison with M51 and M52.

You can notice the jankyness on the page while scrolling on sites like cnn.com, wired.com, amazon.com etc.

 

We initially found the problem when running Chrome M51 and M52 on an Amazon Fire Tablet, and noticed regression across all smoothness metrics particularly high in jank_count(Please find attached file for details).

 

Thanks for the bug details. I will add more details to the bug you mentioned.

 

Thanks,

Sarath | TPM, AWS Silk

 

 

ChromeV52-V51.html
Reply all
Reply to author
Forward
0 new messages