Just detecting WebVR causes constant ~13% CPU usage even when doing nothing else...
Categories
(Core :: WebVR, defect, P2)
Tracking
()
People
(Reporter: mail, Unassigned)
References
Details
User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:65.0) Gecko/20100101 Firefox/65.0
Steps to reproduce:
On a WebVR-capable system (Oculus Rift, HTC-Vive) just doing this simple call causes constant ~13% CPU need:
if (navigator.getVRDisplays)
{
navigator.getVRDisplays().then( function(){} );
}
Here a test page:
https://krpano.com/firefox/bugs/webvr_idle_cpu_need/
Look into the task-manager after opening that page.
Tested on Windows 7 and Windows 10 using Firefox 65 with Oculus Rift and HTC-Vive.
Reporter | ||
Comment 1•6 years ago
|
||
A few more notes:
-
After closing the tab of a WebVR page, it takes a few seconds until the CPU usage goes back to 0% again.
-
This affects all pages that are detecting or using WebVR.
Comment 2•6 years ago
|
||
I guess it might have been resolved in Beta and Nightly at Bug 1514417. Could you take a try to see if there is still the same result in Beta?
Thanks.
Reporter | ||
Comment 3•6 years ago
|
||
Hi, no sorry, that seems to be a different problem - I've just tested Nightly 67.0a1 (2019-02-13) (64-bit) and it behaves the same as in Firefox 65.0.1 - 13% CPU need on my machine just by calling getVRDisplays and nothing else...
Comment 4•6 years ago
|
||
We have just landed a fix for Bug 1537550, which caused some overhead with IPC on every vsync that would be triggered after getVRDisplays. I am hopeful that this has made some impact on your test case.
Reporter | ||
Comment 5•6 years ago
|
||
Thanks!
Is that fix already in the currently Nightly build from today?
I've just tested Nightly 68.0a1 (2019-04-02) (64-Bit) and the bug is still there in that version... (but will test tomorrow again of course).
Reporter | ||
Comment 6•6 years ago
|
||
Tested now the next Nightly build, but the problem is still there.
Btw - it seems that this happens on EVERY system, even without a VR system installed.
I've tested this now also on an older MacMini, there the CPU usage goes up to 20% while doing nothing. Maybe interesting: the OSX Activity Monitor shows an increase of the 'Idle Wake Ups' to ~2000 while the page is open.
Just as reminder you can test and see that bug anytime here:
https://krpano.com/firefox/bugs/webvr_idle_cpu_need/
View the page-source, that's basically a one-liner, only a call to getVRDisplays.
Comment 7•6 years ago
|
||
Some major refactoring affecting this is happening in Bug 1473402. I'd suggest re-visiting after that lands.
Comment 8•6 years ago
|
||
(In reply to :kip (Kearwood Gilbert) from comment #7)
Some major refactoring affecting this is happening in Bug 1473402. I'd suggest re-visiting after that lands.
Bug 1473402 and Bug 1466702 have landed, including a re-implementation of most of VRManager and its related classes.
Could you please check if this has made an improvement?
Reporter | ||
Comment 9•6 years ago
|
||
Hi, tested now the current Nightly - version 69.0a1 (2019-07-03) (64-Bit) - but there it is still not fixed...
Should that fixes already be included there?
Updated•2 years ago
|
Description
•