PowerVR drivers for kernel 2.6 available - 3D acceleration for the N8x0
 

Hi people, digging through the TI website I've found this link http://focus.ti.com/general/docs/wtb...ontentId=27458
So it seems there is now a driver for the PowerVR 3D accelerator with 2.6.x kernel available. I've downloaded it an it has some header files and stuff (didn't have much time to look around).
I don't have enough experience to try to integrate this with the current kernel, but someone used to the kernel src might want to take a look at it.

Hope this helps get the 3D working on the tablet (while I wait for my n800 to travel from USA to Argentina :) )
Please excuse my english, I'm not a native speaker.

cheers to all

Re: 3D acceleration for the N8x0
 

Not possible. I just downloaded it and took a look. They give binary modules for 2.6.14.7-omap2, and OS2008 is running 2.6.21-omap1. Someone find the sources for either that kernel or module, and we might be good to go.

Re: 3D acceleration for the N8x0
 

The 2.6.x version is only for 2430 AFAICT. NITs use 2420 processors. Besides they seem to be for Montavista Linux - that is usually a sign that something is closed.

Re: 3D acceleration for the N8x0
 

AFAICT the OMAP 2430 has a PowerVR MBX Lite, a subset of 2420's PowerVR MBX, so the drivers should work http://www.arm.com/community/display...=2099&display=.
If someone want to give it a try, I'll be glad to help.
One question, why is it that the IT is using a -omap1 kernel instead of -omap2 that is the kernel patchset for the IT's processor?

Re: 3D acceleration for the N8x0
 

I'm giving it a try, will report back how I get on.

See my comments on irc:

http://mg.pov.lt/maemo-irclog/%23mae...11-14.log.html
http://mg.pov.lt/maemo-irclog/%23mae...11-15.log.html

Re: 3D acceleration for the N8x0
 

Promising work lardman, I'll meet you on IRC later today.

Re: 3D acceleration for the N8x0
 

I didn't want to open a new thread about 3d acceleration, so I used the search function and this one seemed the most suitable one for the argument...
I stumbled across this site http://www.htcclassaction.org/respon...ate_20080214_2, people complained about HTC not releasing a good driver for their devices and they finally got an update: why on earth can't we do the same thing? I mean that should be our right to have the best support for the hardware we paid for, I've seen many people on this forum and in the maemo community working hard to create or port new apps or games for the NITs, but I can't say nokia worked as hard as they did... Many flaws in nokia's NIT's can't be corrected (N8x0 slow video interface, for example) and we have to live with them, but the things that these devices could do if only there was a serious commitment by nokia are so many that we can't just sit here quietly: think about the 2d/3d chip integrated in the omap, think about the missing java... Nokia seems to be experimenting with our money, I think it's time we get something in return...

Ps: Maybe I sound a bit on the revolutionary side, but reading every day the gaming section of this forum and discovering that there are many among us NIT user who port games just to discover that they are slow for the missing 2d/3d acceleration really grinds my gears...

Re: 3D acceleration for the N8x0
 

Technical limitations.

Re: 3D acceleration for the N8x0
 

What do you mean? Is it about the "slow" serial connection between the OMAP and the display? Or the N8x0 has the 2D/3D accelerator and the imaging video accelerator cores hardware disabled? Or it's not possible because they work just with the on-chip display controller?
I read everything on the forums and on the maemo community about 3D acceleration, but I still haven't found an official nokia statement about the possibility to have this feature enabled: wouldn't it be at least polite if nokia let its customers know something more? I'd really like if nokia engineers wrote something about these supposed technical limitations in a consistent way, I'm an engineer myself, so it would be easyer for me to understand some technical paper instead of reading "maybe in the future" or "could it be" on NITs related websites...
Obviously if something like that has been already done please could you link those documents/discussions to me?
Thanks and sorry for the bothering, but reading about the HTC class action made me hope for something more from Nokia

Re: 3D acceleration for the N8x0
 

The video hardware cannot handle fullscreen updates with a high framerate and you will always get tearing. But I think pixel doubling still would be a good option for games. Can the 3D accelerator be used with pixel doubling? If yes, then Nokia should really get TI to port the PowerVR driver to Kernel 2.6.

Re: 3D acceleration for the N8x0
 

I refuse to think that having 2d acceleration enabled can't give us better and faster scrolling on full screen web browsing of complex webpages (I'm not talking about 800x480@30fps video or videogames, I know that is impossibile due to the LCD interface): not having it is a real pain, sometimes vertical scrolling is so blocky it remembers me my first 486 laptop...

Re: 3D acceleration for the N8x0
 

Yeah, that's the problem, but we can't help you with that :-)

AFAIK there is no 2d acceleration on the epson chip. To scroll page one pixel up you need to transfer whole frame to the epson chip again. There is scaling or pixel doubling(=2x scaling) with intepolation, color space conversion, rotation, but everything is applied on the fly to rectangle sent to the chip. There is no operation that can move pixels already transferred to the epson chip (=scrolling).

Datasheets here
http://lists.maemo.org/pipermail/mae...ry/014656.html

Re: 3D acceleration for the N8x0
 

Yes, It could. But we don't have simpler things than that. For this to work you would need support for different resolutions and bit depth and its switching in x server and framebufer driver. It is not there. However there is one 640x480 video plane allocated just for video output that can be overlaid over normal screen so maybe it could be reused for such 3d acceleration window. But it is a bit hackish solution and not something they will hurry to do for us. And maybe it is even not possible since this plane is in on-chip OMAP SRAM (AFAIK) not ordinary SDRAM so maybe PowerVR MBX would not be able to draw to this plane. But maybe even yes and SRAM is better for this, I don't know. Just wanted to say that there are many problems to solve.

Re: 3D acceleration for the N8x0
 

:mad: that wasn't very kind... actually it was no light offence, but I guess that you're fanoush and as long as I use your initfs to boot from SD I can take it :p


I just read all the discussion, you never stop to learn something new :)
About the Epson controller I'm pretty shocked about its performances (or the lack of), I'll read better the datasheet later, but it seems there's not much to be happy for :(
Just one last question: on the E90 they use 2d/3d acceleration, the screen is almost the same size as n8x0, the cpu is the omap2140, so, due to hardware limitations in screen resolution did they use an external chip for graphics or did they stick with the integrated accelerator and just used a better interface than the S1D13745?
Thanks

Re: 3D acceleration for the N8x0
 

Oh come on, Nick, don't take offense at fanoush's harmless joke-- that's actually mild for this forum. Stick around. You'll see. ;)

Re: 3D acceleration for the N8x0
 

:D I was just kidding :D

Re: 3D acceleration for the N8x0
 

Than sorry about that. It was just a joke. And a slight reminder that nothing is as easy as it looks so one really has to think before drawing conclusion. 2D acceleration of OMAP2 chip means nothing for us since the framebuffer is external and every change needs to be transferred there before it appears on screen.
Well it is just LCD display controller, not full GPU. It is meant for displaying video and targeted for phones and for that the functionality it is pretty good. I guess they used it because they are phone company so they built first 770 from spare parts they already had and then just kept the design in N8x0. AFAIK they could not use just the display controller built to OMAP1710 due to large display and the shared memory architecture. The LCD refresh DMA (=reading 800x480x16bits 60 times per second) would eat lot of memory bus bandwidth, slowing down everything inside OMAP. As for OMAP2420 the on-chip SRAM is still too small to hold whole 800x480 so there would be same problem. There are no public docs for OMAP2 chips but the TI page
http://focus.ti.com/general/docs/wtb...emplateId=6123
says "5-Mb internal SRAM" (b stands for bits) = 640KB, good enough for 640x480x16bits VGA mode but not good enough for 800x480 (=768KB). Hopefully with new generation tablets the external controller will be gone.

Re: 3D acceleration for the N8x0
 

Just wanted to say thanks for the most interesting and informative thread on this forum for a long time.

Re: 3D acceleration for the N8x0
 

It's informative, so I rated it 5 stars.. and the way the rating system (currently) works this thread now shows 5 stars even though there's only myself rating it (so far) :D

Re: 3D acceleration for the N8x0
 

Fanoush mine was a joke too, you don't have to worry, np ;)
About the omap2 I was aware of the resolution-limit for the built-in display interface, the informations I really missed were the ones you linked in the previous post on the maemo community plus an in-depth knowledge of the epson interface... now I see why nokia doesn't provide this kind of support...
Thanks again for the help :)

Re: 3D acceleration for the N8x0
 

Yarr.... just re-read it all. Very interesting stuff.