[Koying]

Now that 3.10 is up to beta I hope the manufactures all start will start moving to it too.

Are there updated libs somewhere?

[emveepee]

I've see the freescale master

http://git.freescale.com/git/cgit.cgi/im...1.0.0_beta

and since SolidRun seems to based on BoundaryDevice

https://github.com/boundarydevices/linux...1.0.0_beta

Martin

[Koying]

I've see the freescale master

http://git.freescale.com/git/cgit.cgi/im...1.0.0_beta

and since SolidRun seems to based on BoundaryDevice

https://github.com/boundarydevices/linux...1.0.0_beta

Martin

We were speaking about the VIVANTE GPU libs, not the kernel

[emveepee]

We were speaking about the VIVANTE GPU libs, not the kernel

Maybe it is all here https://plus.google.com/+JonNettleton/posts/PJm24HodU4V now.

@stephan, should we continue to use your utilite kernel patches with the new xbmc-im6 repo I am having problems, with overscan correction with calibrate, the x,y axis seems to move for the video playback too, and I end up a green colour segment where there is no video at the top. I was thinking this might be because of the kernel patch

General though the repo itself is going from good to great everyday. I 'don't know if it is partly GeexBox defaults but I do have these other observations

- HDMI is default which is good, but the default is pass through audio
- If HDMI is not on when powered up, HDMI audio cannot be configured without a restart

Thanks for the great work to you all.

Martin

[Koying]

Do you know who's doing the GeexBox builds?
As they probably are the most popular xbmc imx6 builds, it would be good to coordinate with them...

[mijanek]

I have a question only partialy related to xbmc.. when running xbmc on my matrix, I'm getting time to time stalled CPU messages. Is this happening to you too, or am I missing some kernel patch?

Code:

INFO: rcu_bh_state detected stall on CPU 1 (t=0 jiffies)

[mtx512]

I've see the freescale master

http://git.freescale.com/git/cgit.cgi/im...1.0.0_beta

and since SolidRun seems to based on BoundaryDevice

https://github.com/boundarydevices/linux...1.0.0_beta

Martin

We were speaking about the VIVANTE GPU libs, not the kernel

The 3.10.17_1.0.0_beta GPU libaries are available from Yocto (master-next), but they require a kernel patched with p13 vivante changes.

[wolfgar]

We were speaking about the VIVANTE GPU libs, not the kernel

Maybe it is all here https://plus.google.com/+JonNettleton/posts/PJm24HodU4V now.

@stephan, should we continue to use your utilite kernel patches with the new xbmc-im6 repo I am having problems, with overscan correction with calibrate, the x,y axis seems to move for the video playback too, and I end up a green colour segment where there is no video at the top. I was thinking this might be because of the kernel patch

Hi Martin
Regarding my patches, I have to check formally. But I would say yes globally (IPU prio is still interesting, so are workaround for 24bits samples and fix for flickering line of deinterlacing...)


Stephan

---

Do you know who's doing the GeexBox builds?
As they probably are the most popular xbmc imx6 builds, it would be good to coordinate with them...

Hi

It is already the case. I sync regularly with them
The 2 active people for imx6 geexbox are Rudi (you already know him) and Thomas...

Regards
Stephan

[rico]

Hi, first of all, thanks for the good work porting xbmc to imx6. I think we are really close to getting a good XBMC experience on ARM devices.

I've got a few questions.

The i.mx6 SOLO has one VPU, the i.mx6 DualLite and higher have two. Does that make any difference for the performance, or does the current xbmc imx6 port use one VPU as of now anyway?

And if that doesn't make a difference, then a i.mx6 quad wouldn't be faster with xbmc then an i.mx6 SOLO, since xbmc does not use the extra cores, right? (assuming you aren't running any other heavy services)

[wolfgar]

Hi rico,

You are right the number of cores will not make a big difference for xbmc
Yet there are other differences between solo and dual/quad. Especially the DDR bus width (32bits vs 64bits) and frequency are not the same (at the end memory bandwidth should be better for imx6q). Also the GPU are not the same
These differences can matter for xbmc
The best would be to perform exhaustive tests for given use cases on both platform and check how it behaves on the 2 different socs...
For now I cannot tell that the final experience will be exactly the same for sure but the solo behaves well given its price

Stephan

[Koying]

I only have a quad, but assuming the imx-specific code works ok on solo, I'd guess the experience will be close to the one of an RPi.
Still guessing, but the differences between a Dual and a Quad should barely be noticeable from an XBMC pov.

It would be very interesting to do tests on a dual-lite...

[jurafxp]

Is it possible to decode Hi10P AVC videos on i.MX6?

[CruX]

The 3.10.17_1.0.0_beta GPU libaries are available from Yocto (master-next), but they require a kernel patched with p13 vivante changes.

I've been running those binaries in combination with wolfgars 3.0.35 kernel plus the mentioned patches for quite some time.
No problems so far!

I have not made the change to the actual 3.10.17 kernel yet though.

Cheers

[abreaks]

I am having a halting problem on an Arch based version of XMBC for my Wandboard Quad. The halting only effects XMBC. I can continue to work through SSH. The log file shows this error over and over again:

ERROR Decode - not hw buffer available. Waiting for 2ms


I can't find anything useful through Google on how I can fix this. Can anyone here point me in the right direction?

[wolfgar]

Hi abreaks,

does it happen all the time or only with specific streams ?