[xmarsx]

Great that works, thank you very much

Is there someone with write access to the repository to fix this 'round' issue on ARM? Otherwise a bug report + patch file would be great.

I assume you are using the Linaro groups Ubuntu 11.10 builds for Pandaboard correct?

[davilla]

Is there someone with write access to the repository to fix this 'round' issue on ARM? Otherwise a bug report + patch file would be great.

I assume you are using the Linaro groups Ubuntu 11.10 builds for Pandaboard correct?

It's not a bug but a feature I suspect that if you pass "--enable-neon" into configure, then it will be fine. As I've mentioned below, how neon is treated changed recently. Pandaboard does have armv7 with neon.

[naj666]

Is there someone with write access to the repository to fix this 'round' issue on ARM? Otherwise a bug report + patch file would be great.

I assume you are using the Linaro groups Ubuntu 11.10 builds for Pandaboard correct?

No I am using the latest images from the ubuntu site for armel, they are recently update to the newest linux kernel 3.0. This also fixes the problem not able to set the screen resolution to a lower setting making the pandaboard and xmbc performing well for video's as the hardware acceleration for 1080p is not in place yet.

[xmarsx]

@naj666 and @davilla
Thank you for the comments.
I'm going to give the Linaro build of Ubuntu for Pandaboard a shot (http://www.linaro.org/downloads/11-10/).

[naj666]

also, xbmc's configure was changed with respect to neon, needs --enable-neon passed in to turn on neon.

I can confirm that the --enable-neon also works, tested it on a clean xbmc clone.

[rako2911]

How is the pinout to connect via TTL?
Thanks

[naj666]

Dirty regions seams not to be working on pandaboard, the screen becomes black and lights up where ever the mouse is.

[davilla]

How is the pinout to connect via TTL?
Thanks

You connect TX to TX and RX to RX, don't forget GND, and PWR depends on your serial-to-ttl device.

[McGeagh]

It's not a bug but a feature I suspect that if you pass "--enable-neon" into configure, then it will be fine. As I've mentioned below, how neon is treated changed recently. Pandaboard does have armv7 with neon.

Useful bit of info:
Tegra2 is the only SoC that is armv7-a and without NEON (that i know of).
(NEON is only optional for CortexA9's, but Tegra2 is the only one which opted out.)

[xmarsx]

Hello,
I was actually successful building xbmc (pvr) for the Pandaboard on Ubuntu 11.10 using the TI SGX libraries.
After installing the packages mentioned in the README.ubuntu file (plus the TI 'libgles*-sgx-omap4*' and 'libegl1-sgx-omap4* files) I called:

Code:

./bootstrap
./configure --enable-gles --enable-neon --disable-debug
make -j2

I don't know which other parameters make sense to be used on the Pandaboard.

Running the 'make -j2' took 3 hours 30 minutes with a lot of pseudo warnings:

Code:

the mangling of 'va_list' has changed in GCC 4.4

I still have to do more tests as this morning I only started the build and called different menu items before going to work.

As some might be interested in this here is the full 'odyssey' getting a Ubuntu 11.10.

Originally I wanted to build XBMC it on the Linaro Ubuntu 11.10 builds for the Pandaboard but there is currently a problem generating the system for the SD-Card (at least from a Ubuntu 10.04 LTS Desktop system) because there is a failure adding the 'ti-uim' stuff (Linaro Bug #886101) with the linaro-image-tools. On a Desktop System with Ubuntu 11.10 and maybe 11.04 it should work.

I then tried the preinstalled Linaro Ubuntu Pandaboard image (with hardware accelerated graphics driver) which is ready for SD-Card deployment. I had some major display errors with this one. If anyone here wants to test it get the preinstalled Linaro Ubuntu 11.10 Desktop package here (packages for other boards like e.g. BeagleBoard, Origen, etc. are also available at this location):
http://releases.linaro.org/images/11.10/...u-desktop/

For installation of the image you can also use the command described here at the Ubuntu Image for ARM page using zcat and dd on one line:
https://wiki.ubuntu.com/ARM/OmapNetbook

As the above Linaro build had some problems I finally got to the official Ubuntu 11.10 Desktop release for armel omap4. The installation procedure wasn't that stable as it crashed several times after I selected the german keyboard layout but finally I got through it (this is no overheating issue as my Pandabaord has a big headsink attached; max temperature compiling XBMC was 47 °C).

Then I installed the TI graphics drivers (ubuntu-omap4-extras-graphics) as well as the multimedia packages (ubuntu-omap4-extras-multimedia - because of a conflict the metapackage had to be deselected again but the already selected packages for the multimedia parts could be installed) for Ubuntu 11.10 (aka Oneiric Ocelot):
https://launchpad.net/~tiomap-dev/+archi...er=oneiric

With the multimedia package (gstreamer and gst) and graphics driver installed I wasn't able to play videos with Totem nor VLC. I could hear some sound but generally the system just hangs itself instantly.
Clarification: Totem 3.x directly crashes when I try to play an XviD encoded AVI file (haven't tried an *.mvk with h.264 yet). VLC uses it's own code for decoding (not gstreamer) and plays the video on the X desktop in a window (TOP tells me it uses ~34% of CPU). When I hit the Stop button of VLC the X desktop crashes and is restarted. When I try to play the XviD fullscreen it becomes a slideshow (lots of skipped frames).

So far I've not yet tried it again without those packages as at least some hardware acceleration would be nice.

Update 2011-11-15: With the latest TI update released yesterday the crashes trying to play a video file seem to be gone at least with Totem 3.x (TI-packages from ppa: ubuntu-omap4-extras - 1.1 - 2011-11-14, pvr-omap4 - 1.7.9.2.1.7-1 - 2011-11-09, and so on). Playing an XviD video at 1:1 sized window the CPU usage is at 15% but there are some screen glitches at the lower half of the screen (it think I read this in a bug description...seems to be a known but still unresolved problem). The glitches also show up on fullscreen and it looks as if some packages are dropped but it's nowhere near the previous slideshow.
VLC as well as XBMC tests follow shortly.
Update 2011-11-20: xbmc is playing XviD (SD) content pretty well. Currently it's only too slow when you play HD content. Maybe I take a video and put it on Youtube when I get the time.
VLC on the other hand wasn't running as good as Totem, but I think that could have been expected. It seems it's not using the acceleration support offered by the TI components.

[BrokenCodes]

Useful bit of info:
Tegra2 is the only SoC that is armv7-a and without NEON (that i know of).
(NEON is only optional for CortexA9's, but Tegra2 is the only one which opted out.)

No, actually, the Trident line of SOC's have not NEON, and this line of SOC's is in the products that I support with their current crappy closed source middleware...
It is SoftFP only...
I can post the specs, but for the most part they are a normal Cortex A9 with PowerVR SGX 53x.

[McGeagh]

No, actually, the Trident line of SOC's have not NEON, and this line of SOC's is in the products that I support with their current crappy closed source middleware...
It is SoftFP only...
I can post the specs, but for the most part they are a normal Cortex A9 with PowerVR SGX 53x.

Interesting. Thanks for that. Not heard of Trident before! Cheers

[robclark]

Any fourther steps in aplying hardware acceleration with XBMC on the pandaboard yet? I'm running Ubuntu 11.10 now and there is still no way to get 1080p movies playing with more than 1/5 fps Even with the rep (https://launchpad.net/~tiomap-dev/+archi...er=oneiric)..

Hey, btw, I happen to know a bit about hw decode and gfx/display on omap4xxx.. I wrote libdce and gst-ducati, done some work on xorg driver, working on drm kernel driver for omap, etc. So maybe I could help in some way. It would be really cool to have xbmc running fully hw accelerated on panda, and could be an interesting side project..

Anyways, was just starting to poke around the xbmc src tree and try to get an understanding for how the video decode/rendering works, and what might be needed.

I see on topfs2's github tree there is some gstreamer decoder support. I guess that should be sufficient for decoding? Or is there any reason that adding a native decoder would be better? There are some things like cropping codec output buffer (removing codec-edges) that you might loose out on w/ appsink approach. Cropping comes as an downstream-event from decoder to video sink, normally. Also, ffmpegcolorspace isn't really the hot ticket for performance.

What about on the rendering end of things? Has anybody looked at the IMG texture streaming extensions? Or alternatively, I've been working on an extension to dri2 to support video. It might be interesting to have a dri2 renderer. The idea with dri2 video would be that your OSD is a separate drawable (itself potentially using dri2) which the xorg driver can blend with the video however it sees fit (ie. use GPU, or ARGB blending in display with hw overlay).

Anyways, I'm a complete newb to the xbmc code base, just started looking at it this morning. But I know the hardware, so maybe with a bit of point in the right direction I could be helpful..

[Cofi]

I have been trying to get Linaro to work with the pandaboard the last few days.
Like xmarsx said above there are some graphic errors in the Linaro Ubuntu build and gfx freezes at startup.
So instead i tried the ALIP version. Boots fine, but after installation of OMAP4 extras and reboot it refuses to start. I managed to run XBMC without the OMAP4 extras at one point but it was really slow.

[xmarsx]

I have been trying to get Linaro to work with the pandaboard the last few days.
Like xmarsx said above there are some graphic errors in the Linaro Ubuntu build and gfx freezes at startup.
So instead i tried the ALIP version. Boots fine, but after installation of OMAP4 extras and reboot it refuses to start. I managed to run XBMC without the OMAP4 extras at one point but it was really slow.

It might be worth trying newer Linaro builds. Maybe the desktop graphic errors are now fixed (nightly builds)...as TI released newer components last week.

As I also wrote above on the standard Ubuntu Desktop/Server release there were some screen issues in the lower half of the screen when playing videos with Totem (only then).
xbmc on the other hand is running perfectly (XviD,..) on the standard Ubuntu 11.10 as long as you don't try play HD content which is a little bit slow. For HD xbmc has to use the available hardware of the Pandaboard better.
Maybe robclark can jump in here. From what he wrote about libdce and gst-ducati he seems to be the right person for this task.