Hacker News Reader

Comment by antouank | Hacker News Reader

> Additionally we'll ship an accelerated rendering pipeline for Linux/GNOME/X11 users for the first time, ever!

I'm seeing high CPU usage on YouTube after recent updates on Xubuntu. I think it started when they enabled GPU acceleration by default.

Previously I could watch even FHD videos without much fan noise, but after recent updates, fans spin at the max speed by playing even 480p quality vids. Anyone else?

GPU acceleration is not a magic bullet it seems. In our case drawing large SVGs suddenly becomes really slow when GPU rasterization is enabled. Chrome has the same problem. I really hope we will see more polishing in this area and maybe a cooperation between Chrome and Firefox developers?

Thing is, when I play the same video with VLC which uses HW decoding, it doesn't even break a sweat. So I think there might be something wrong on the browser part...

You can issue a bug report on https://bugzilla.mozilla.org/home Sometimes bugs are fixed really fast, sometimes never. Just give it a try.

There could be something related to FAN thresholds that resides in BIOS directly thus starting using GPU just slightly over some minimal limit might raise FAN speeds significantly and not be related to temperature raise as there is some inertia between temperature readings and real temperature inside GPU core.

I'm also experiencing FAN noise raise on my Ryzen APU based laptop.

i mostly stopped bothering with in-browser video when mpv wiht youtube-dl integration came around. either drag the url to the player window or use an open-with addon.

hardware decoding, no ads, buffering as i please

Do you have an example url for this?

Here is a bug report that shows the issue on Chrome. It is roughly 7 times slower with hardware rasterization enabled:

https://bugs.chromium.org/p/chromium/issues/detail?id=114367...

This particular example is only slow on Chrome. Even if you move the edge of the shape into the viewport Firefox is much faster than Chrome. But I know from real life examples that Firefox becomes slow if handling large SVGs. I will have to investigate this and file a bug report.

I checked again with our app on Firefox and performance is not that bad. Rendering gets a slower with large SVGs, but it is still good and there is not much difference between hardware rendering enabled and disabled. So I was either wrong about slow performance on Firefox or things have improved recently.

Would you mind capturing a profile? https://blog.paul.cx/post/profiling-firefox-media-workloads/... explains how.

It's very very very simple and quick, there is a video demo and a text explaining what to do, and then you can open a bug at https://bugzilla.mozilla.org/enter_bug.cgi?product=Core&comp..., with the profile link (you can login with a github account or email). This will be triaged and hopefully we'll understand what's up. Please use a Nightly build or a Mozilla-provided build when doing so (so there are symbols). It's just a matter of downloading and untaring the tarball.

Please also attach a copy of `about:support` raw data (button at the top), so we can understand what your configuration is: desktop Linux systems can be pretty diverse and sometimes it isn't clear what works, what doesn't and why.

If you're playing 480p, then youtube might be serving you AV1, and this is (for now) using the CPU (trading a lot less bandwidth for increased CPU usage), you can check the codec it's using by right clicking on the video and clicking on `Stats for nerds`, there is the codec string. You can adjust your codec preferences in your youtube profile I believe.

Just did some tests on Firefox 83 and DOM reflow time grows exponentially with the depth of DOM tree. With grid elements of depth 10 single reflow takes seconds on my Ryzen laptop while on chromium is not noticable at all.

Can you post a link to a test case?

Did already profiling and posted a bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1682686

https://jsfiddle.net/fuh7c894/

I'm on Linux. I stubled on webrender from a blog yesterday, and enabled it. My observation was the same - high CPU usage, so I then disabled. Got 84.0 today, and gave it another go. It appears to be behaving well so far. So I'm guessing there was some fix in 84.0 that made it work for me.

Does anyone know how to enable that?

I'm on Nightly so I should have it for some time already, but I'm not sure if it is on by default or should I tweak some value.

Is this about WebRender? If so, here you go: https://www.omgubuntu.co.uk/2020/07/firefox-enable-webrender...

What does "Compositor:" say for you in about:support?

gfx.webrender.all

layers.omtp.enabled

layers.acceleration.force-enabled

layers.omtp.enabled isn't used when WebRender is on.

ah, nice - thanks

Wait... GNOME only?

Yeah, that seems odd. I wonder if the update doesn't help/involve Weyland people, and if GNOME is an X11 holdout? I haven't been paying enough attention to know.

Perhaps just by default, or a miscommunication. FWIW I’ve had WebRender force enabled on Sway (Wayland i3 clone) since FF 78 or so.

Yep. We've seen bugs with different window managers so WebRender is only enabled in Gnome in 84 to minimize the risk of breakage. This restriction will be removed in subsequent releases.

Also, we have it enabled in nightly and beta for many more configurations already :). There are many bugs to fix, but we're working through them. This is the exact same treatment we gave Windows and OSX in previous releases this year.

I know the usual rational about users blaming Firefox and yadda yadda but I am rather not convinced this is how things should work in the open desktops. If there are clear bugs in other environments, file them and break them if necessary. The only potential exception I can see is when it's not clear if it's a bug or not..

It's not clear, is it only for gnome users ?

Reply on news.ycombinator.com