Intel Has Been Working On A Fast 2D GPU Renderer Focused On Web Content

[timofonic]

Isn't the rust based renderer of servo insanely fast, too? How do they compare?

Furthermore, the GTK+ developers acknowledged, that cairo is slow for GPU accelerated rendering and thus, they are working on a replacement.

What do they want to replace it for? What about all software using Cairo?

Does Keith Packard have something to say about this? This is an Intel project and he works for Intel

[Michael]

Does Keith Packard have something to say about this? This is an Intel project and he works for Intel

Keith Packard hasn't worked at Intel in a year and a half.... http://www.phoronix.com/scan.php?pag...d-Leaves-Intel

[liam]

Anyway they are still pretty faster at rasterizing. After turning off the text that made things unequal I got this on my Sandy Bridge desktop

Cairo CPU: 22fps
Cairo GL: 25fps
Qt CPU: 19fps
Qt GL: 58fps
FastUIDraw: 80fps

Not that though I switched to Qt5, it still uses the deprecated QGLWidget API.

Any chance you could through skia into the mix?
I'd imagine that's going to be the fastest of the lot since its got the most developers working on it (and the most users relying on it).

Once again, I'm very disappointed that TyGl seems to have been ignored by all.

https://github.com/szeged/TyGL

[wizard69]

They are probably using all the other toolkits wrong. Using them right would probably make them all 4x to 10x times faster.

When nvidia talked about their accelerated path rendering, they were at least using qt in an absurdly slow and stupid way causing it to be 10x slower than it should be. With Qt used right hardware accelerated paths are only a 2-3x times faster. Still good of course, but I am skeptic of these kinds of benchmarks where the authors have no interest in representing the alternatives in a realistic light.

Both QT and Cairo need modernization and by the way a 2-3X jump in performance is very significant and any off loading of the CPU ends up being a win.

[timofonic]

Keith Packard hasn't worked at Intel in a year and a half.... http://www.phoronix.com/scan.php?pag...d-Leaves-Intel

Oh, thanks. I didn't know!

[carewolf]

Any chance you could through skia into the mix?
I'd imagine that's going to be the fastest of the lot since its got the most developers working on it (and the most users relying on it).

I seriously doubt it. It has the same architecture and limitation as the Cairo and Qt 2D renderers. These are not slow because of incompetence, they have very similar performance, because they are both doing similar work about as fast as you can.

[carewolf]

I seriously doubt it. It has the same architecture and limitation as the Cairo and Qt 2D renderers. These are not slow because of incompetence, they have very similar performance, because they are both doing similar work about as fast as you can.[/SIZE]

But I did want to test it. I didn't because skia is terrible to work with. If there were Debian packages I would have used those, but there aren't, and last time I checked skia out from git and tried to use it, it neither compiled out of the box, nor did it work reliably. I might have been unlucky, but in my mind it is a piece of software that is part of Chromium and it just isn't well-maintained as a stand-alone component.

[hansg]

Both QT and Cairo need modernization and by the way a 2-3X jump in performance is very significant and any off loading of the CPU ends up being a win.

Just curious, what sort of modernisation did you have in mind? I've been using Cairo for a couple of years and I'm (overal) quite happy with it. Which is not to say there isn't room for improvement, of course...

[vw_fan17]

Speaking of optimization and such.. I understand that the end goal is to make a faster web browser. I haven't looked into X11 performance optimization lately - any way to speed up firefox/chrome with an nvidia gtx 950 and nvidia blob? I.e. enabling open-gl or something?

[liam]

I seriously doubt it. It has the same architecture and limitation as the Cairo and Qt 2D renderers. These are not slow because of incompetence, they have very similar performance, because they are both doing similar work about as fast as you can.[/SIZE]

Actually, its arch is quite different from cairo, at least. Not sure about qtpainter (I think its called).
Also, implementation matters a whole lot, and skia tends to benchmark quite well.

Edit: skia is absolutely a fast moving target, which is projects which use it as a standalone, like mozilla, take snapshots and use wrappers; with that in mind, you MIGHT take a look at those wrappers like azure (not sure what its redependencies are or what kind of api it exposes).