![]() |
Re: N900 - Yes, it sucks.
|
Re: N900 - Yes, it sucks.
Quote:
AWESOME. WANT. NOW. HATE. DELL. |
Re: N900 - Yes, it sucks.
Quote:
It's very interesting to see that the issue doesn't occur (or only very slightly) when zoomed out and showing a complex web page with tens of thousands of characters and plenty of images. It only occurs when you've zoomed in fully. That tells me it is definitely a software issue with the rendering algorithm. If it was raw performance, it would be slower when zoomed out, due to increased scene complexity (whether hardware-assisted rendering or not). Since that's Gecko's area (the heart of Mozilla, Firefox, Fennec) I don't think we can blame Nokia for Gecko's behaviour - after all Gecko is an excellent browser engine choice in most respects - but we can annoy the folks at Mozilla with these videos until they improve it :D I have no comment on the other issues. |
Re: N900 - Yes, it sucks.
Quote:
Anyway, this issue is of no concern to me. Even Megacrazy's videos present performance that seems satisfactory to me. And if I want another browser, that shouldn't be a problem either. |
Re: N900 - Yes, it sucks.
Quote:
|
Re: N900 - Yes, it sucks.
Quote:
|
Re: N900 - Yes, it sucks.
I've recently implemented phase 1 of a double-tap zoom hack for the Diablo GTK webkit, so I spent some time staring at how the iPhone OS webkit implements zoom. It's simple. They just do an animated 2D transform while another thread is rerendering the page. They time the animation such that the final pixel copy happens soon after the animation finishes. If you have such a device, you'll notice that your screen is frozen to UI while this is happening. After the zoom is complete, you have to reregister your onFingerDown event. Nothing will happen until you do. A simple parlor trick, nothing more. Nokia could do this, but instead they decided to show you the continuous output of your appSSS. Plural. So, now all you eye candy weenies know.
|
Re: N900 - Yes, it sucks.
Quote:
In fact it has the delay twice there. It really looks like it happens when there are certain object types being rendered. See what I mean about it only happening when zoomed in? And I notice, when scrolling up (to the start of the page); it doesn't happen scrolling down. At 4:18-4:20 when you zoom out, there's a long delay in filling in the top and sides - more white gap. When zoomed out, it's pretty consistently good. At 4:12, there's a white strip at the top of the browser - another rendering delay, perhaps. (It'd be less visible if it was black). (I see lots of brief stalls in the kinetic scrolling, but that could easily be Youtube not the device). Big rendering chug at 5:33. Ok, that's a task switch, we may expect more chugging then. (Though a really polished implementation would hide such things). Quote:
As someone who has worked on game engines, video playback devices and GUI toolkits, I tend to notice these little things and think "if only I it wasn't just me..." :p I'm fascinated that you didn't see them in your own video. It really reminds me of my colleague who can't tell the difference between 30fps games and 60fps games - whereas I can tell immediately and they feel totally different to look at. No wonder people argue about these things if they're invisible - and therefore not an issue - to half the people :D |
Re: N900 - Yes, it sucks.
Quote:
|
Re: N900 - Yes, it sucks.
Quote:
I think Nokia's approach is better, and in this case it's up to the app to provide the parlour tricks. (Though for really polished polish, to make transitions glitch free you do need some global scene management or compositing of app output.) |
All times are GMT. The time now is 15:07. |
vBulletin® Version 3.8.8