•   
  • 1
  • 50
  • 51
  • 52(current)
  • 53
  • 54
  • 104
  •   
Allwinner A10 : Is XBMC ported to MALI-400MP ?
(2012-08-15, 09:48)johanjonker Wrote:
(2012-08-15, 00:40)Nu7s Wrote: In my opinion, Allwinner has already done more then most large companies (Samsung, LG, ...) would. I'm very curious how the above mentioned files will turn out. Really hope XBMC on my Mele A2000 will be possible. It's a great piece of hardware for an awesome price.

Why would you say Allwinner has done more? they havent been doing much, there is already 2 tv top android devices that is getting full support between the XBMC team and the chip manufacuter, the Ouya and Pivos. Now I just want to point out that the Ouya is boosting a Tegra 3 quad core for $99..

It's also still a little way away from actually making it to market yet, mind.
Reply
The Ouya will only be available in a couple of months. Now think, how much progress is made every month? By the time the Ouya will be available, there will be something better in the make that will release in a few months from then. And so the circle continues... The A10 is available now at a cheaper price than the Ouya will be in a couple of months.

And by "doing more" I meant they have replied to the request from the XBMC community to have a new look into their libraries and work togheter with the XBMC team. Maybe this co-operation hasn't been very intensive, but they did something.

If they succeed in helping to bring XBMC to the A10 they will realise that working in the same direction for their upcomming models pays off. And who knows, maybe a new model will be released in the same time as the Ouya, with comparable hardware for a better price and with native XBMC "support".

Don't get me wrong, I love the Ouya, great idea, great people behind it, i'm a backer. But it's not the holy grail. And nobody knows what the future will bring when it releases.

Reply
(2012-08-13, 23:40)Shivansps Wrote:
(2012-08-13, 14:20)gimli Wrote:
(2012-08-13, 10:44)Krankdroid Wrote: so. still no news about the Allwinner communications ?

No. The communication with Allwinner is neraly zero.

Well Allwinner should be long working on A10 reeplacement by now, i dont think they care much about it...

I know I sound like one of those nagging evangelists, but if they would have helped or at least released programming specifications early on with some basic examples, someone for sure would have picked up on that. Further support would then not be required. Maybe some QA back and forth. A binary blob can be ok, to have initial 'licenced' support.

They are probably indeed working on sun5i now and probably don't care much anymore. Hopefully they'll learn from this. As many said, a10 is missing out on some potential here.

That all said, as linked above, https://github.com/iainb/CedarXWrapper seems to be what is required to reverse engineer the driver and get an opensource video decoder going. So the first steps have been made. Let's hope this all comes around before sun6i comes out Smile

As for ouya etc, tegra3 support is even worse I belive at the moment?
Reply
(2012-08-15, 12:55)oliv3r Wrote: As for ouya etc, tegra3 support is even worse I belive at the moment?

Tegra 1/2 doesn't support NEON instructions for the ARM processor, but Tegra 3 does, so Tegra 3 should be all good for running XBMC.
Reply
(2012-08-15, 13:08)Ned Scott Wrote:
(2012-08-15, 12:55)oliv3r Wrote: As for ouya etc, tegra3 support is even worse I belive at the moment?

Tegra 1/2 doesn't support NEON instructions for the ARM processor, but Tegra 3 does, so Tegra 3 should be all good for running XBMC.

But only using binary blobs I assume. As far as I know, nVidia is still behind in submitting kernel patches for the tegra3, I think it was over at phoronix that tegra3 support wasn't all roses and moonshine.
Reply
(2012-08-14, 03:04)Boon_nz Wrote: According to Eva from Allwinner "Yes. We are still discussing with Gimli for a solution. "

I wouldn't call it a discussion.
Reply
What was the real email address from Eva ? Please PM it to me.
Reply
(2012-08-15, 13:08)Ned Scott Wrote: Tegra 1/2 doesn't support NEON instructions for the ARM processor, but Tegra 3 does, so Tegra 3 should be all good for running XBMC.

Hi,
I asked some things about NEON in another thread (http://forum.xbmc.org/showthread.php?tid...pid1166442), but I have not received any answer yet.
I would like to understand how NEON is working. Is it for accelerating all kinds of codecs? Strong enough for decoding h264 1080p videos? e.g. in a 1Ghz arm cpu? or is it for something else?
Thanks in advance
Reply
(2012-08-15, 18:02)pszab Wrote:
(2012-08-15, 13:08)Ned Scott Wrote: Tegra 1/2 doesn't support NEON instructions for the ARM processor, but Tegra 3 does, so Tegra 3 should be all good for running XBMC.

Hi,
I asked some things about NEON in another thread (http://forum.xbmc.org/showthread.php?tid...pid1166442), but I have not received any answer yet.
I would like to understand how NEON is working. Is it for accelerating all kinds of codecs? Strong enough for decoding h264 1080p videos? e.g. in a 1Ghz arm cpu? or is it for something else?
Thanks in advance

NEON is a vector math processor, like mmx or sse in x86 world. It does nothing by itself, you have to write code to use it.
Reply
(2012-08-15, 19:58)davilla Wrote:
(2012-08-15, 18:02)pszab Wrote:
(2012-08-15, 13:08)Ned Scott Wrote: Tegra 1/2 doesn't support NEON instructions for the ARM processor, but Tegra 3 does, so Tegra 3 should be all good for running XBMC.

Hi,
I asked some things about NEON in another thread (http://forum.xbmc.org/showthread.php?tid...pid1166442), but I have not received any answer yet.
I would like to understand how NEON is working. Is it for accelerating all kinds of codecs? Strong enough for decoding h264 1080p videos? e.g. in a 1Ghz arm cpu? or is it for something else?
Thanks in advance

NEON is a vector math processor, like mmx or sse in x86 world. It does nothing by itself, you have to write code to use it.

Yes, I know. And, can the written code be stong enough to play hi quality videos compared to a simple arm cpu code which is very poor?
I'd like to know that can we state that if neon is present, the actual chips (~1Ghz) can play everything (with a good code) and we can avoid situation like in a10 case... or other manufacturer specific vpu accelerations are needed...
Reply
over simplification of NEON. for example, an h264 decoder written in NEON is very non-trivial. If you start now, then maybe in a year it might be done.
Reply
(2012-08-16, 03:16)davilla Wrote: over simplification of NEON. for example, an h264 decoder written in NEON is very non-trivial. If you start now, then maybe in a year it might be done.

Well, you say there aren`t any written NEON decoders yet. I thought there are, because I saw some full-neon and non-neon called apks.
What are you using for currentl if nor for decoding?
Reply
Hello there, my first post from spain.
Has anyone send to gimgli eva's email?
Reply
(2012-08-16, 08:52)pszab Wrote:
(2012-08-16, 03:16)davilla Wrote: over simplification of NEON. for example, an h264 decoder written in NEON is very non-trivial. If you start now, then maybe in a year it might be done.

Well, you say there aren`t any written NEON decoders yet. I thought there are, because I saw some full-neon and non-neon called apks.
What are you using for currentl if nor for decoding?

Correct, there are not any pure NEON video decoder except what we get from FFMpeg.

Hardware video decode is done by dedicated video decoders provided on the arm SOCs. Arm SOC vendors use different IP designs so any API is not common to all devices. One must use a higher level API such as OpenMax.

I think you are confusing the need for NEON with hardware video decoders. Given a fast enough ARM device with multiple cores, video decode can be handled purely in software (FFMpeg), this is evident from reports from tegra3 devices being able to decode 720p HD content. But software decode costs CPU and more CPU means greater power consumption. For example, the Nexus 7 doing 720p HD content seems to saturate two cores, while the same content on an Amlogic based device idles at less than 10 percent.

NEON give the headroom to do all the other things that XBMC does. Like generate a thumbnail in a timely fashion. No NEON and this simple task can take 5 to 10 seconds or more. With NEON, less than a second.

Reply
(2012-08-16, 17:15)davilla Wrote: Correct, there are not any pure NEON video decoder except what we get from FFMpeg.

Hardware video decode is done by dedicated video decoders provided on the arm SOCs. Arm SOC vendors use different IP designs so any API is not common to all devices. One must use a higher level API such as OpenMax.

I think you are confusing the need for NEON with hardware video decoders. Given a fast enough ARM device with multiple cores, video decode can be handled purely in software (FFMpeg), this is evident from reports from tegra3 devices being able to decode 720p HD content. But software decode costs CPU and more CPU means greater power consumption. For example, the Nexus 7 doing 720p HD content seems to saturate two cores, while the same content on an Amlogic based device idles at less than 10 percent.

NEON give the headroom to do all the other things that XBMC does. Like generate a thumbnail in a timely fashion. No NEON and this simple task can take 5 to 10 seconds or more. With NEON, less than a second.

I thought Neon could be an alternative to accelerate video decoding and instead of using vendor APIs...
Thanks for the clarification!
Reply
  •   
  • 1
  • 50
  • 51
  • 52(current)
  • 53
  • 54
  • 104
  •   
 
Thread Rating:
  • 15 Vote(s) - 4.13 Average



Logout Mark Read Team Forum Stats Members Help
Allwinner A10 : Is XBMC ported to MALI-400MP ?4.1315