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

Issue 729166 link

Starred by 1 user

Issue metadata

Status: Untriaged
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: 2017-06-06
OS: Android
Pri: 2
Type: Bug

Blocked on:
issue 729730



Sign in to add a comment

Nexus 5X power.idle_platform shows strange power spike at the beginning of the trace

Project Member Reported by charliea@chromium.org, Jun 2 2017

Issue description

For some reason, the power is always much higher at the beginning of the power.idle_platform stories than at the end. The length of this high-power period seems to vary from story to story. This variation is a source of noise in the Android power benchmarks.

It seems that many systrace events during this period are things like "MicrophoneReader", "MicroRecognition", and "AudioIn_52".

This seems to happen in every trace. Here are a few examples:

https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_0-2017-06-02_05-23-50-58957.html

https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_2-2017-06-02_05-24-06-68069.html

These events continue for a while after Telemetry's RunStory method begins.

I vaguely remember a conversation about "Okay Google" detection being a source of noise in the lab. sullivan@, rnephew@, nednguyen@: do any of you know what might be going on here?
 
Cc: ccraik@google.com jreck@google.com
+Android experts: I was guessing that it's ok Google's voice recognition, but not sure that's correct. You have any idea how to stop this background process?
I'm interested not just in how to stop the background process, but what we might be doing differently than users that's causing the background process to trigger in the first place. I'd prefer to do the latter, but will settle for the former if it's something that's a really hairy problem.
I'm wondering if its the screen turning off... we disable charging so if there is no interaction with the phone, the screen will turn off. Is there a way to look for that in the atrace data?
The NextAction date has arrived: 2017-06-06
Blockedon: 729730

Sign in to add a comment