RenderingNG: Ready for the next generation of web content

[lionkor]

I'm probably just being cynical, but those graphs and diagrams scream of a lack of "meat" to what they're saying. Except for the first graphic, they are all just products of a marketing brain, as far as I can tell.

Showing the old engine as a graph that plummets, and the new one as one that doesnt, or the one that plots "frustration" against "features & complexity", is just utterly meaningless to me. Maybe this appeals to the small part of the community that gets excited for quirky and cool posts by big corps, but I only see about 5 paragraphs of info here.

Basically RenderingNG will more reliable and better and better and more reliable, and thats neat, I just wish they didnt waste so much of the reader's time.

[roca]

I think the Chrome team has done a lot of good work here, but you're right about the graphs. What units does the "frustration" axis use, anyway?

[egnehots]

Yeah, but don't forget that's just the first post in a serie.

Let's hope that the next post will have more meaningful technical content..

[peakaboo]

Look at the evolution of the web technologies. Lots of improvements. Now look at the web. It's unusable without an ad blocker. The tech mafia keeps tracking people and censoring the web, practicing right-speak and removing critical voices. Can a new Web technology solve that? If not, it doesnt matter.

[contravariant]

Clearly in the old version the scrolling performance becomes negative at some point, which would be interesting to see.

[Ristovski]

Rendering performance is one of the few things left that Chrome has up its sleeve as an edge over their competitors (mainly Firefox), at least on Linux.

Out of the box, Firefox has terrible rendering performance, not to mention hw-accelerated video decoding. Now, Chrome is not a whole lot better out of the box, since it still applies old "Driver bug workarounds" that have long been fixed in mesa/GPU drivers, but this is easy to circumvent with the `--disable-gpu-driver-bug-workarounds`.

I've been running Chromium with the said flag for over half a decade now, and I have yet to see one of the bugs manifest. Firefox on the other hand, has a similar entry in `about:config`, but one needs to tinker with even more flags to get Firefox to acceptable performance (Somehow, enabling xrender makes WebGL fast, but makes video decoding have weird jitter, etc).

The day Firefox gets comparable WebGL/video decode performance will most likely be the day I switch.

[floatboth]

> one needs to tinker with even more flags to get Firefox to acceptable performance

All you need is gfx.webrender.all, if you even failed the qualification (most modern setups shouldn't fail it).

[bufferoverflow]

Seriously, FF on Ubuntu causes 100% CPU usage when watching YouTube in 1080p. Meanwhile Chrome on Windows eats maybe 6-10%.

How is this not #1 priority to fix?

[roca]

Personally I just have layers.acceleration.force-enabled: true and that's enough to get me fast 4K WebGL. WebRender and Wayland dmabuf seem to turn on automatically. Intel 915, Fedora 34, dual 4K monitors, distro Firefox 89.

[AHTERIX5000]

Yeah it's not great on Mac either. I have pretty much standard Firefox installation with only Ublock Origin added and it's extremely easy to get 16" Macbook Pro hot when Safari handles the same load without too much heat.

[tormeh]

Is this Chrome catching up to Firefox's Quantum projects? I.e. Stylo and WebRender? It puzzles me a bit that Firefox, Chrome, etc. don't reuse each others components. Obviously Chrome can't just plop in WebRender, but is it really more effort to integrate it than to write your own? Same for Firefox and V8. I'm not saying all browsers should be the same, but (and this is a big assumption) if reuse is possible, then wouldn't it make more sense to focus on your team's strengths and use external components for the rest?

[flohofwoe]

The article reads more like its a pat on the shoulder for general cleanup work on the Chrome render pipeline that's already been happening since 2014 and that's now nearing the finish line. Also exchange of implementation ideas is more important than exchange of code, especially for implementations of web standards. Microsoft reusing Chromium is already bad enough for the web as a whole (as a counter example, AFAIK the WebGPU implementation teams have been working closely together, but each on their own implementation, which sounds pretty much perfect to me, healthy competition, but without reinventing the wheel).

[roca]

Using external components is pretty hard. The interfaces between them are pretty complicated and abstraction layers are expensive. There are also different policies in different browsers, e.g. Firefox is much more into using Rust than Chrome is, for obvious reasons.

Firefox does use Skia though.

[readflaggedcomm]

>Gecko and Webkit have also implemented most of the same architectural features described in these blog posts, and in some cases even added them before Chromium.

That's why all this threaded rendering stuff sounded familiar. Mozilla putting massively-parallel Servo features into Firefox is already a few years old: https://en.wikipedia.org/wiki/Gecko_(software)#Components

[pverghese]

Yet for some reason. Firefox is still slow compared to chrome

[emersion]

A little disappointed to not see Linux mentioned in any of the timelines.

[d_tr]

Indeed. Chromium feels very fast on Windows with a Ryzen 2600 and an R9 280X GPU (4 TFLOPS). On Linux it feels sluggish and a single zoom step can take up to a full second on some pages, and nothing I tried (Wayland vs X11, amdgpu vs radeon, experimental chromium flags vs defaults) seems to make any difference...

[stolenmerch]

Am I correct in reading that these NG sites aren't accessible without a side DOM synced to canvas content?

[ReactiveJelly]

They're saying it's possible, but the article is talking (mostly fluff) about some software improvements to Chromium to enable that.

I don't know if there's such a thing as an "NG site", and if there is, the new software will still support normal sites

[meibo]

> You should not need to worry about browser bugs making features unreliable, or breaking your site's rendering.

As long as Safari is still around to haunt us, I think I will :)