<aholler>
[ 253.819] (EE) AIGLX: reverting to software rendering
<aholler>
(and I was wrong yesterday, dri is only provided as binary, the drv is what I compile myself)
<aholler>
anyway, at least I have now a starting x11. even if it is slow.
<aholler>
next step, testing xbmc with patches for imx and directfb
<aholler>
you see, I've no idea about what I deal with. I did some OpenGL programming several years ago but since then I kept away from display-stuff. ;)
<malik>
why do you want x11 for xbmc?
<malik>
it runs fine without it, probably better
<aholler>
don't want it, just haven't found a way to start it without
<malik>
i compiled xbmc-imx on my imx6 and it ran fine
<malik>
a lot of people do it
<malik>
completely without x11
<aholler>
maybe, I'm totally new to xbmc. Haven't done much more with it than checking out a git-version and called xbmc ;)
<aholler>
and it said it wants x, even if I've configured it without
<malik>
i do not have my configure parameters anymore, but you can disable it, definitely
<vpeter>
I think openelec runs xbmc without x. ( I'm working with it on tbs matrix board).
<malik>
i think the important step at configure was enable-gles or something
<aholler>
I did so
<malik>
vpeter: actually i think nobody runs this stuff under x11.... it only makes things harder
<dv_>
x11 also makes it impossible to use vsync
<dv_>
be sure to use "export FB_MULTI_BUFFER=2" before running xbmc
<aholler>
ah
<vpeter>
this will speed up aholler's work :-)
<dv_>
you can achieve vsync in x11, but you need some dri2 functions for that. and dri2 isnt what you get from vivante_dri.so
<aholler>
I've just an existing ebuild in gentoo and build from some imx-tree at github. enabling gles but not X. And it says: http://dpaste.com/3KAQ664
<aholler>
+used
<malik>
aholler: better post your configure results
<aholler>
they are gone. anyway, let me add the patch for directfb and let me look myself. ;)
<malik>
in 3.17 the vpu will be supported in mainline if i see things correctly. then i will throw away xbmc and not use the gpu at all anymore when watchding videos.
<malik>
for me every once in a while now the vivante drivers crash and i have to reboot... and the gpu is used in xbmc only for the menu anyway
<malik>
directfb? why do want this now?
<aholler>
I don't know, I haven't seen a switch to use fb
<malik>
i do not use directfb
<malik>
i doubt anyone else does
<aholler>
so just enabling gles should work without X11?
<malik>
i had a lot of configure options but basically egl + gles, yes
<malik>
no x11 at all
<aholler>
will look at it again, thanks for the info
<aholler>
malik: just using gplay or similiar would be nice too. Also xbmc has a nice menu and cec is nice to use too
<jnettlet>
malik, for the mx6 branch of xbmc the gpu is used for displaying the video as well.
<malik>
thinking about it... i think the ipu can do scaling and the ipu csc can do format conversion, right? i don't know how well this is supported in software anywhere
<dv_>
I support it in gstreamer-imx
<dv_>
the IPU can be picky though
<malik>
jnettlet: hm. i do not get this. how does the gpu make the syncing problem easier?
<dv_>
certain operations are not supported, and it is not fully clear what the criteria are
rz2k has joined #imx6-dev
<dv_>
if you do rotation, you cant do it with surfaces larger than 1024x1024. but non-rotated scaling and csc are supported even with 1080p
<jnettlet>
malik, because then you let the vpu just decode frames as fast as it can and the gpu can worry about blitting them over to the framebuffer at the proper time
<jnettlet>
if the VPU has to also worry about moving frames around at the proper time it will bottleneck the decoding stream
<dv_>
yeah
<malik>
jnettlet: interesting.
<malik>
dv_: interesting as well...
<malik>
good channel to learn something obviously :)
<dv_>
deinterlacing is also possible with the ipu
<dv_>
also, keep in mind that the ipu can do csc,scaling,rotation,deinterlacing in one step. no need to do this in multiple ones
<dv_>
but as always, the software side is a bit rough. there is an "ipu-lib" floating around the web, which is not usable for the imx6. instead, for the imx6, you have to use kernel headers
<malik>
dv_: is the ipu driver in mainline supporting this stuff in any way?
<dv_>
tbh, I'd rather use G2D instead of the IPU if G2D had deinterlacing support. the API is considerably cleaner. there is also PXP, which I know nothing about yet
<aholler>
hmm, I've seen g2d for 3.10.31 ;)
<aholler>
whatever that is
<dv_>
g2d is the vivante gpu's 2d core
<jnettlet>
pxp is much simpler. It is basically for handling camera image sensor data and making it suitable for use by the display or vpu
<dv_>
it is generally better to use it (or the IPU, or PXP) instead of the 3d core for 2d content
<dv_>
because the 3d core uses much more power
<dv_>
jnettlet: could pxp be used for a gstreamer transform element like the IPU?
<dv_>
and/or as a video sink (that is, blitting frames to the framebuffer) ?
<jnettlet>
there is nothing stopping it from being used. I think it is exposed to userspace via v4l2
<dv_>
it would be interesting to have a pro/con table for IPU/G2D/PXP
<jnettlet>
oh know it exposes an IOCTL interface to the dma engine part of it
<jnettlet>
I have never looked at the userspace side of it :)
<dv_>
I've decided to improve gstreamer-imx' ipusink a bit, since many people had concerns about imxeglvivsink's power consumption
<jnettlet>
let me know what you find
<aholler>
I wouldn't trust the dma. still have to disable it on the uart for uploading the bcm-firmware ;)
<jnettlet>
hmmm, never experienced any problems with it.
<aholler>
brcm_patchram_plus hangs here when I've enabled dma on the uart.
<malik>
i had problems with crappy freescale kernel and uart dma on mx25 and mx5 but not on mx6 iirc
<malik>
but not with any other dma
<aholler>
can't say about what else use dma, but everything else seems to work (besides the graphics I'm still playing with)
<aholler>
and the problem doesn't happen always, but often enough to render bt unusable with dma on the used uart
<aholler>
something in 3 out of 4 calls to brcm_patchram_plus just do hang. might be a race somewhere.
<aholler>
that happens especially after a reboot
<aholler>
(warm-boot)
<aholler>
so it might be a reset-problem too
shogunx has quit [Ping timeout: 276 seconds]
Er0l has joined #imx6-dev
shogunx has joined #imx6-dev
Er0l_ has quit [*.net *.split]
Er0l_ has joined #imx6-dev
Er0l_ has quit [Ping timeout: 244 seconds]
kroon has quit [Quit: Leaving]
sjoerd has joined #imx6-dev
paulk-collins has joined #imx6-dev
<aholler>
ah, progress. building xbmc failed when I've switched the imx-libs to x11. After switching the libs to fb build was ok. ;)
<aholler>
(still not tested as I'm rebuilding with some other things changed) but at least it gives me hint that it might have failed because I've used the wrong imx-libs. And the remote git changed too (around 200 patches more, likely rebased)
stunpix has quit [Ping timeout: 260 seconds]
steeve has quit [Remote host closed the connection]
RaYmAn has quit [Ping timeout: 240 seconds]
RaYmAn has joined #imx6-dev
RaYmAn is now known as Guest73228
tumdedum has quit [Ping timeout: 260 seconds]
tumdedum has joined #imx6-dev
tumdedum has quit [Ping timeout: 255 seconds]
tumdedum has joined #imx6-dev
diego_r has quit [Quit: Konversation terminated!]
tumdedum has quit [Read error: Connection reset by peer]
tumdedum has joined #imx6-dev
jas-hacks has joined #imx6-dev
steeve has joined #imx6-dev
steeve has quit [Remote host closed the connection]
jas-hacks has quit [Remote host closed the connection]
obinou_ has quit [Remote host closed the connection]