Re:Mobile Firefox: Measuring how a browser feels
by pfraze on Tuesday June 26, 2012 @11:30AM
(#4162404)
Speaking of Mozilla mobile, make sure you see this talk [1] on their recent product designs. Don't know if this article's android browser is the same as the ios browser in the talk, but the ios browser looks great.
Re:Mobile Firefox: Measuring how a browser feels
by ender7 on Tuesday June 26, 2012 @01:32PM
(#4163076)
This is a really good step forward, but it's disappointing to see the author using FPS to measure UI responsiveness (I've harped on this before: http://news.ycombinator.com/item?id=3777058).
While FPS is a decent measure of performance in constant-load applications such as the fishbowl, it's really, really bad for sporadic-load applications such as most UIs.
For measuring UI responsiveness, you should really use frame delay, which is the greatest elapsed time between any two concurrent frames during a UI event. If your swipe gesture has a frame delay above ~33ms, it will look bad, even if your measured framerate is ~30fps.
Re:Mobile Firefox: Measuring how a browser feels
by Retric on Tuesday June 26, 2012 @02:07PM
(#4163310)
One of the big problems with frame delay is it tends to round up. On a 60FPS device 30.1FPS generally means you miss one out of every 2 refreshes. But a 35ms frame delay and you missed 2 refreshes and can even lose 3 which is ~100ms and rather noticeable.
Re:Mobile Firefox: Measuring how a browser feels
by ender7 on Tuesday June 26, 2012 @03:03PM
(#4163611)
True, but having 30.1FPS doesn't protect you from that phenomenon either. If you maintain 30.1FPS over a long period of time, it means that, yes, you generally get one out of every 2 refreshes.
The problem is that UI interaction events can take place over very short stretches -- say ~200ms. That's only a handful of frames, which reduces the quality of any metric that relies of averages (such as FPS). I don't really care if I get 1 out of every 2 refreshes as measured over 2 or 3 minutes, I care about the first 5 frames after I swipe the screen.
Re:Mobile Firefox: Measuring how a browser feels
by densh on Tuesday June 26, 2012 @01:37PM
(#4163114)
I don't know if it's a good place to ask this question but what happend to Firefox for Android tablets? I hasn't been updated for some time. Can developers share any details about that?
Re:Mobile Firefox: Measuring how a browser feels
by AaronMT on Tuesday June 26, 2012 @03:43PM
(#4163772)
Right.
We currently don't support tablets in this release as development did not finish in time for this '1.0' style release.
Fortunately, you can test out Nightly (our developer oriented release channel) which has some tablet-optimized changes available for download at http://nightly.mozilla.org
[I'm a Mozilla QA Engineer]
[1] https://air.mozilla.org/product-design-at-mozilla/
I am talking about the Android platform
While FPS is a decent measure of performance in constant-load applications such as the fishbowl, it's really, really bad for sporadic-load applications such as most UIs.
For measuring UI responsiveness, you should really use frame delay, which is the greatest elapsed time between any two concurrent frames during a UI event. If your swipe gesture has a frame delay above ~33ms, it will look bad, even if your measured framerate is ~30fps.
The problem is that UI interaction events can take place over very short stretches -- say ~200ms. That's only a handful of frames, which reduces the quality of any metric that relies of averages (such as FPS). I don't really care if I get 1 out of every 2 refreshes as measured over 2 or 3 minutes, I care about the first 5 frames after I swipe the screen.
We currently don't support tablets in this release as development did not finish in time for this '1.0' style release. Fortunately, you can test out Nightly (our developer oriented release channel) which has some tablet-optimized changes available for download at http://nightly.mozilla.org [I'm a Mozilla QA Engineer]