Firefox 4 Feature Directory

[mernen]

Okay, there's something weird here:

These are my results for Microsoft's Psychedelic Testing and FishIE Tank, running in full screen at 1280x1024:
Windows 7, 32-bit build, render-mode = 0: 2 RPM; 250 fish at ~4 fps, 50 fish at ~14 fps, 10 fish at ~17 fps
Windows 7, 32-bit build, render-mode = -1: 1779 RPM; 250 fish at ~52 fps, anything below that at a solid 60 fps
Ubuntu 10.04, 64-bit build: 1779 RPM; 250 fish at ~9 fps fps, 50 fish at ~35 fps, anything below at 60 fps

Many other demos show similar patterns, sometimes completely indistinguishable from the D2D version.

As you can see, the Linux build is not performing as well as WIndows' D2D, but still much better than when render-mode is set to 0. Maybe there's already partial support?

[KWierso]

Render-mode 0 is D2D OFF.
Render-mode -1 is D2D ON.

It makes sense that it performs better with render-mode -1.

[Omega X]

Take it to the Direct2D thread. This is not the place for D2D troubleshooting.

[El Pino]

Are you sure hardware acceleration is not yet supported in Linux? Lately Firefox has been rendering ridiculously fast here, beating Microsoft's own IE9 HTML and canvas demos (SVG is still slow), even with large, maximized/fullscreen windows. I just can't see that happening without hardware acceleration.

Today this post showed on planet.mozilla.org: http://blog.mozilla.com/cjones/2010/08/ ... -on-linux/ So there is some hardware acceleration on Linux as well, apparently. And on some pages that makes a massive difference indeed.

[Omega X]

Any of the regular *nix users want to verify that OpenGL is indeed working for regular content?

I didn't see any standout csets that enabled this.

[mernen]

Render-mode 0 is D2D OFF.
Render-mode -1 is D2D ON.

It makes sense that it performs better with render-mode -1.

I think you didn't understand the purpose of those lines. I was just comparing them to the Linux build, to show it stands somewhere inbetween. (By the way, the Linux build seems to ignore render-mode altogether, but this isn't too surprising)

Today this post showed on planet.mozilla.org: http://blog.mozilla.com/cjones/2010/08/ ... -on-linux/ So there is some hardware acceleration on Linux as well, apparently. And on some pages that makes a massive difference indeed.

Yes, that's exactly what I'm talking about! (Though it didn't start just today, my first post was written in the 20100818 build. I'm fairly sure at least the 20100817 build was like that too. I didn't test Microsoft's demos, but other typical HTML pages felt really really lightweight)

Take it to the Direct2D thread. This is not the place for D2D troubleshooting.

The D2D part was just used as a reference for comparing the Linux performance. Anyway, I'm sorry for starting all this discussion here; originally, I thought it'd be done in one or two posts, someone confirming the acceleration in Linux. It seems to be somewhat of an unknown (probably caused by a specific interaction between other components, like Cairo and X), and my misunderstanding of the state of D2D just made things worse.

I guess we should stop deviating here, and only post on this topic if some solid information is found. In case anyone thinks this is a discussion worthy of a new topic, I'm willing to participate.

[Omega X]

MSI Packaging officially got the boot since it won't make the freeze in time.

[BigWillyStyle42]

Any of the regular *nix users want to verify that OpenGL is indeed working for regular content?

It's not OpenGL acceleration done by Firefox; it's the X server doing the acceleration. This is not the same as the D2D acceleration for Windows.

[Omega X]

Without getting into a drawn out semantics tirade about what is doing what in Firefox.

I only need an answer to the question at hand.

[BigWillyStyle42]

Without getting into a drawn out semantics tirade about what is doing what in Firefox.

I only need an answer to the question at hand.

Very well, the short answer is "no".

[avada]

Can anyone tell me what happened with the alt+d hotkey? I doesn't seem to be working.

[Omega X]

Wrong thread for that. Try creating a new thread or asking in the support forum.

[mat--]

Chromium's video scaling code port just landed on trunk (see https://bugzilla.mozilla.org/show_bug.cgi?id=577843).

[Omega X]

Added.

Also I changed Per-Document Garbage Collection to Compartmental Garbage Collection. It seems to be the same thing. But If its not, feel free to let me know.

[sabret00the]

Looks like the Account Manager will make it in before the freeze is Gavin Sharp can clear his review queue in time.