[LINUX] XBMC for Linux port to ARM architecture CPU and SoC chips? - Printable Version
+- XBMC Community Forum (http://forum.xbmc.org)
+-- Forum: Development (/forumdisplay.php?fid=32)
+--- Forum: Feature Suggestions (/forumdisplay.php?fid=9)
+--- Thread: [LINUX] XBMC for Linux port to ARM architecture CPU and SoC chips? (/showthread.php?tid=35139)
[LINUX] XBMC for Linux port to ARM architecture CPU and SoC chips? - Soju - 2008-07-23 17:55
I've been following the progress of the other ports (PPC, Cell), and wonder if porting to a higher end ARM/PowerVR is at all possible?
This device, that became available recently, has a Freescale MPC5121e SoC, which is minimally OpenGl 1.0 compatible. In theory PowerVR (Series 5) can do OpenGL 2.0.
- althekiller - 2008-07-23 19:07
Not any time soon. That microprocessor is no where near fast enough to run XBMC, let alone with linux beneath it (yes, I'm aware that ARM are more efficient clock for clock than x86). I have my doubts that there are any kernel drivers available to fully expose full opengl 2.0 for that GPU. Not to mention you'd need someone who actually wants to do the porting of not only xbmc, but most of its dependencies as well.
Also, AFAIK no one has begun porting to cell. A couple guys are (were?) working on PPC linux. There was rumor of someone porting to cell, but we haven't seen any code to my knowledge, so I'm very skeptical on that front.
- Gamester17 - 2008-07-23 19:44
Already explained here (same thing):
Anything is possible but I doubt anyone on Team-XBMC is interested in porting XBMC to ARM.
- Soju - 2008-07-23 19:49
Thanks for the replies. IMHO there is a drastic difference between sigmadesign chips and ARM 11 cores with a higher end PowerVR implementation or better. So the dependencies are OpenGL 2.0 and processing power? At what point does ARM become interesting? Would a 1 Ghz dual core Arm 11 suffice to run 720p, assuming there is OpenGl 2.0 support on the GPU? There is also plethora of Linux support for Arm 9 and 11. Lastly, the price point is phenomenal and the MIB market will make them cheaper still over time. The little PC I mention is going to be one of many.
- Gamester17 - 2008-07-23 21:03
Soju Wrote:Would a 1 Ghz dual core Arm 11 suffice to run 720p, assuming there is OpenGl 2.0 support on the GPU?Probably not 720p video encoded with H.264 codec, no.
FYI; the OpenGL requirement of the GPU is for the XBMC GUI, (not much at all of the actual video-decoding is of-loaded to the GPU, the CPU practically have to take the whole load of the video-decoding process by itself).
I do not think it is interesting because no one of Team-XBMC developers owns one of these machines them self and as XBMC is a hobby project (with everyone on Team-XBMC working on XBMC for free in their spare time only for the fun of it) they all only code things that they have an personal interest in.
You are most welcome to start porting XBMC to ARM yourself and submit patches to the XBMC project, I am sure that everyone on Team-XBMC would do their best to try to support you in the effort (as much as we can anyway without having access to the hardware), ...which is the same as the deal with XBMC on PowerPC and Cell (which the difference that there are probably a whole lot more people out there who already own or have access to that hardware).
- Soju - 2008-07-23 22:04
Hmm. I actually do have some access to coders , although I'm not myself one (I do work in tech sector professionally). The thing is, I'm trying to identify an ARM SoC platform that will at least in theory can be successful. Based on both of your replies, it seems there are a few serious dependencies:
1. Processing power
-Arm11 vs x86 performance per core, per cycle.
2. Future GPU codec/etc off-loading schemes
-Whether XBMC advancement in offloading processing to GPUs in the future will be generic enough to work on PowerVR varients, the usual GPU ARM coprocessor.
-Specific case of Nvidia (rather than PowerVR) Purevideo decoders becoming available for Linux generally, and Arm bytecode specifically. At present Tegra only supports MS OSs- which may or not present a target for the windows XBMC as a potential target as well.
3. Picking a Linux distro,
-potentially the MIB Ubuntu- as a candidate OS, so it has general compatibilities with the rest of the Linux effort for XBMC linux port.
If anyone is interested in joining the effort, perhaps I can acquire and contribute a target device, once identified, for it.
- Gamester17 - 2008-07-23 23:01
hmmm, ...why does it have to be of ARM processor-architecture?, why not just instead go for Intel's Atom CPU (and Atom Centrino chiptset), VIA Nano (Isaiah), or the upcoming competing low-energy x86-platform product from AMD (codename "Bobcat")?, (all of those are x86 and compatible with XBMC without modification).
I mean, what are the minimum requirements you are looking for and at what 'raw' hardware material price per unit?
- althekiller - 2008-07-24 00:32
XBMC would already run on the atom given the system had the rest of our requirements. Advantages of ARM/MIPS are super low power/heat/price while still performing quite well.
- Soju - 2008-07-24 01:09
With the Nvidia entry, the price performance will only improve. Then there is size, this thing is a SoC- it produces way less heat, and takes in way less juice. The platform has potential, but perhaps not just yet as a prepacked device, for Home theater. But even if there where bare-bone options, if a build was trivial a few of us will jump at it.
Please go right ahead and make it happen if you got the resources... - Gamester17 - 2008-07-24 10:43
Soju Wrote:If anyone is interested in joining the effort, perhaps I can acquire and contribute a target device, once identified, for it.Feel free to promote and lobby this as much as you want, ...the more platforms XBMC runs on the better IMHO. So please go right ahead and make it happen if you got the resources to do it (ie. skilled C++ programmers and cash).
Note though that I personally do not think there is a market for such hardware unless your company or other companies built and sold pre-built set-top-boxes with XBMC for Linux (or for Windows XP Embedded) on it pre-installed and working out-of-the-box. I feel that for someone to pay more than $200-300(US) for that it would have to be capable of playing back 1080p H.264 encoded videos (meaning full-bitrate Blu-ray movies rips in Full HD), ...if it would only capable of 720p H.264 encoded video playback then I do not think many people would be willing to pay more than $199(US) at the most, and it would still have to be a XBMC pre-installed set-top-box that work out-of-the-box.
Again, you should checkout Intel's Atom CPU (Silverthorne) and Atom Centrino chipset (Poulsbo) which is an passivly cooled x86 platform that does all of that at less at 5-10W, see:
- EsOsO - 2008-08-17 20:05
Hi Soju, any updates on this project? It sounds good!!
What about this:
- Soju - 2008-08-18 20:26
Thats a nice board. I'm mulling this over (waiting to see how the Via Nano/Intel Atom make a splash and if the Nvidia Tegra platform is successful and is ported to Linux). I think I agree with Gamester17 in that, at least initially, choosing an appropriate HW platform and specializing for it is the way, and if one is picked it pretty much has to play decent hi-bit 1080P video. The competition like Popcorn Hours latest standalone MB is pretty intense. W/O extremely optimized drivers I'm not sure that todays Arm11 Platforms will be able to do it w/ Linux-XBMC. Do you think otherwise?
- EsOsO - 2008-08-18 21:09
Hi Soju, i agree with you about choosing the right HW platform and now it seems a bit premature cause HD decoding isn't already embedded in GPU.
My goal is to achieve a xbox-like platform with a reasonable boot-time, low power consumption and fanless design.
I'm not an hardware guy so i'm in your hand about capabilities of Arm architecture but i can keep an eye on what's happen in this world.
I'll be glad to help you in some way cause i feel that this "project" could bring xbmc to the next step!
- Gamester17 - 2008-08-19 14:42
if you (or anyone else) own a company that would like to go for a ARM processor for the commercial purpose of making a profit mass-producing XBMC based then you should maybe think about hiring a skilled C++ programmer that could code a DirectFB renderer for XBMC to use instead of OpenGL (just like how Team-XBMC developers made an OpenGL hardware graphic acceleration renderer to replace the Direct3D we have on the Xbox), my point is that you do not specificly need to use OpenGL but you will need some kind of hardware graphic acceleration that is fast enough to render XBMC's GUI, see:
Like for exampe the PNX8935 based STB225 reference board
or even just use Neuros OSD2.0 for development (thought it does not support 1080p)
- CyruzDraxs - 2008-08-24 22:38
I am planning on converting XBMC to use OpenGL ES 2.0 to see if I can get it running on the Beagleboard with Angstrom. I may, at a later point try to convert it to DirectFB to make it more efficient, but I have no experience with DirectFB so it will likely be awhile.
Please don't expect any releases to come from this, as this is a 'when I have the time' project and I am currently rather lacking in free time.
For anyone who cares; the OMAP3530 in the Beagleboard could almost certainly be able to handle 1080p H.264 if optimized for DirectFB and DSP decoding. It'd be a tight squeeze though.