|
|
2007-12-05
, 15:13
|
|
|
Posts: 3,524 |
Thanked: 2,958 times |
Joined on Oct 2007
@ Delta Quadrant
|
#12
|
The limited frame buffer size is probably because the chip was designed for use with a smaller LCD screen than what's installed in the N8xx. It's a pity, but (as I was about to say, then I noticed the Capt'n used the exact same words!).. there's always a way!
|
|
2007-12-05
, 15:38
|
|
Posts: 2,102 |
Thanked: 1,309 times |
Joined on Sep 2006
|
#13
|
|
|
2007-12-05
, 16:43
|
|
|
Posts: 4,930 |
Thanked: 2,272 times |
Joined on Oct 2007
|
#14
|
That's such a strange limitation. Currently there's a framebuffer capable of holding the entire screen. Perhaps the rendering can be done in two 800x240 halves, and then blitted to memory. Not ideal, as it would incur losses, but the framerate may be acceptable enough for 3D animation.
It may even be possible to render an 8 bit depth. With a proper palette, dithering, and the N8xx pixel density, the resultant 3D may be adequately convincing.
|
|
2007-12-05
, 17:44
|
|
|
Posts: 201 |
Thanked: 88 times |
Joined on Aug 2007
@ San Francisco, CA
|
#15
|
|
|
2007-12-05
, 17:55
|
|
Posts: 29 |
Thanked: 2 times |
Joined on Nov 2007
|
#16
|
|
|
2007-12-05
, 17:56
|
|
|
Posts: 3,524 |
Thanked: 2,958 times |
Joined on Oct 2007
@ Delta Quadrant
|
#17
|
Possible, but I think in practice simply reducing the dynamic screen area would be better. The whole video controller bandwidth issue limits framerates at 800x480; by reducing the size transferred, 30 FPS should be workable.
I don't think this will work; I'm no expert, but I don't know of any 3D accelerated stuff done in a paletted video mode. I think it has to render in RGB. You could do 8 bits of RGB, with 2-3 bits per color, but that would look awful. Dithering works by rendering in high color depth, then replacing a color with a pattern at low depth. The (potential) issue here is requiring us to render into a 5Mbit buffer, so we would not have high-depth data to reduce.

as lardman pointed out, this may be no issue at all. It's not well understood at present, at least by me.
|
|
2007-12-05
, 18:01
|
|
|
Posts: 3,524 |
Thanked: 2,958 times |
Joined on Oct 2007
@ Delta Quadrant
|
#18
|
I think the point here, if you read the thread that was linked, is that the PowerVR is supposed to do its processing in internal memory (internal to the PowerVR chip, and I don't think this is the OMAP's 5Mbit framebuffer). Therefore, it doesn't matter that the OMAP's framebuffer is only 5Mbit as it's not used.
The PowerVR is mapped via kernel memory, so presumably there should be no limit other than the inherent bandwidth limit of the link to the lcd controller.
|
|
2008-03-04
, 22:49
|
|
Posts: 53 |
Thanked: 44 times |
Joined on Feb 2008
|
#19
|
)
N800/OS2007|N900/Maemo5
-- Metalayer-crawler delenda est.
-- Current state: Fed up with everything MeeGo.