<JMann>
Hello, has anyone had any luck with the pcie on the i.MX6?
DV__ has quit [Read error: Connection reset by peer]
dv_ has joined #imx6-dev
aholler_ has joined #imx6-dev
aholler has quit [Ping timeout: 259 seconds]
<JMann>
MIGHT & MAGIC: HEROES VI
JMann has quit [Quit: Leaving]
hste_ has quit [Ping timeout: 272 seconds]
hste_ has joined #imx6-dev
jnettlet has quit [Ping timeout: 260 seconds]
jnettlet has joined #imx6-dev
tasslehoff has joined #imx6-dev
jnettlet has quit [Ping timeout: 255 seconds]
FelixH has joined #imx6-dev
aholler_ is now known as aholler
stunpix has joined #imx6-dev
jnettlet has joined #imx6-dev
jnettlet has quit [Read error: Connection reset by peer]
jnettlet has joined #imx6-dev
rz2k has joined #imx6-dev
obinou has quit [Remote host closed the connection]
obinou has joined #imx6-dev
tasslehoff has quit [Quit: WeeChat 0.4.2]
kroon has joined #imx6-dev
tasslehoff has joined #imx6-dev
steeve has quit [Ping timeout: 260 seconds]
rz2k has quit [Read error: Connection reset by peer]
diego_r has joined #imx6-dev
steeve has joined #imx6-dev
tasslehoff has quit [Quit: WeeChat 0.4.2]
stunpix has quit [Ping timeout: 264 seconds]
_whitelogger__ has joined #imx6-dev
bfederau_ has joined #imx6-dev
tkryger_ has joined #imx6-dev
tumdedum has joined #imx6-dev
tumdedum has quit [Changing host]
awakecoding has joined #imx6-dev
staylor has joined #imx6-dev
diego_r has quit [Ping timeout: 260 seconds]
ragesoss has joined #imx6-dev
kroon has quit [Quit: Leaving]
ragesoss_ has joined #imx6-dev
ragesoss has quit [Ping timeout: 272 seconds]
ragesoss_ is now known as ragesoss
paulk-collins has joined #imx6-dev
FelixH has quit [Quit: Leaving]
<dv_>
aholler: did you find a solution to your problems?
<aholler>
not searched, but at least one problem was solved, I had a stale old lib lying around
<aholler>
but 3.10.31 currently seems unusable.
<aholler>
if I call glxinfo, the driver misses a library call, if I preload libVIVANTE, it misses another call
<aholler>
dv_: I will first switch back to 3.10.17, maybe I will find a usable setup for that. maybe next weekend or such, when I have again some patience to play that ugly setup.
<aholler>
+with
<dv_>
never use glxinfo on arm boards
<dv_>
never use glx or desktop gl on arm boards
<aholler>
never did that too ;)
<dv_>
always use egl and gles
<aholler>
kde with egl?
<dv_>
ah. I see.
<aholler>
usable kde is currently the target ;)
<dv_>
then you are in for a difficult ride
<dv_>
if they use desktop GL, it becomes very painful
<aholler>
you can switch a lot on and off, but I thought thge i.mx has hw-acceleration
<dv_>
it does, but its opengl es
<dv_>
using desktop GL on embedded devices makes no sense. desktop GL is a huge monster with tons of functionality that is not useful and usually not available on embedded GPUs
<dv_>
the real problem is that desktop environments only target desktop GL
<aholler>
I've read some people have glxgears with 300fps. so it must be possible. But I assume they all use a kernel 3.0
<dv_>
you gotta ask yourself a few things:
<dv_>
1) how big was the window when they ran glxgears?
<dv_>
2) what version of the GPU libraries etc. did they use?
<dv_>
3) did they try anything other than glxgears?
<dv_>
vivante drivers arent exactly the best of the world. jnettlet might call that an understatement :)
<aholler>
I don't care. with sw I get 20fps, with hw I got it never running
<aholler>
so the target is to get it running at all
<dv_>
I recommend getting GLES acceleration running first
<dv_>
without that, you can forget about desktop GL
<dv_>
IIRC the vivante libGL lib is actually just a wrapped around GLES
<dv_>
use my eglinfo tool to create a dump about how well the drivers support EGL & GLES
<aholler>
maybe I just will buy a mbm when one with 2gb or 4gb ram will be available
<aholler>
what I already had to patch for 3.10.31 doesn't look very good. Seeing missing prototypes and such doesn't make much courage ;)
<aholler>
s/3.10.31/3.10.17/
<aholler>
(I use c99 to compile the stuff, doing so you'll see a lot more than without) ;)
<dv_>
3.10.31 is a dreaful release
<dv_>
*dreadful
<aholler>
.17 had the missing prototypes already
<dv_>
which is a pity, because very nice improvements are hidden in it
<aholler>
yes, support for xorg 1.15
<aholler>
e.ge
<dv_>
opengl es 3.0 , much better GLES performance
<dv_>
also much better performance with Weston
<dv_>
I wish I could help you more, but I am mainly experienced with yocto
<dv_>
and the situation is much better there
<aholler>
and xbmc seems to need X, at least I haven't seen a fb-only-version
<dv_>
oh, really? thats strange. both xbmc and the viv drivers both support fb-only rendering
<aholler>
hmm, than I have to look how to call xbmc
<aholler>
I just compiled a imx-version I've found and it tells me it needs hw-accelerated X
<dv_>
however, you need some additional cflags when you use fb-only rendering
<dv_>
also, you need some vivante specific calls to get EGL going
<dv_>
actually, imx6 support for the xbmc is under active development. in fact, xbmc has become one benchmark for the cubox-i devices
<aholler>
there is lot of stuff floating around. But I don't want to spend years to fiddle around with pathches for everything and all ;)
<dv_>
you will spend a lot more time trying to go with desktop GL
<dv_>
I have been writing GLES code for the imx6. and also touched GL code. so I can say from first hand experience that using desktop GL there is a big mistake
<dv_>
so, both paths are painful. I happen to think the GLES path is less painful :)
<aholler>
I don't know if I will use it, but it should at least work
<dv_>
GL?
<dv_>
no, no guarantees for that
<aholler>
yes, with segfaults ;)
<aholler>
+out
<dv_>
the vivante libGL has always been rather fragile
<aholler>
besides that, I've compiled kde with egl- and gles-support
<aholler>
but I just haven't played with that. If the xserver crashes at startup, there isn't much to test ;)
<dv_>
can't argue against that
<dv_>
rolling back to 3.10.17 indeed is a plan
<dv_>
retry the egl/gles/kde thing then
<aholler>
first I have to find out which config-switch I changed for 3.10.17. it currently crashes on startup ;)
<aholler>
those kernels are so ugly ;)
<aholler>
even more ugly than those for omap ever were ;)
<aholler>
sorry for the rant, but when I call git describe and it tells me foo+2000 or foo+1200 I don't know what else to say ;)
<aholler>
and running into a bug I've posted a patch 2 years ago doesn't raise mood ;)
<jnettlet>
aholler, the fastest glxgears I have seen is 346fps
<aholler>
the fastest I've seen was 20fps (sw only). But I'm sure I will find all necessary patches sometimes. Maybe I've just missed one when testing 3.10.17
<jnettlet>
but dv_ aholler, KDE's kwin does have a native EGL/GLESv2 renderer
<jnettlet>
aholler, right now with the binary vivante_dri.so that you need to do OpenGL acceleration under X, the limitation is Xorg versions
<jnettlet>
anything newer than 1.14.3 I think and it won't work so well
<aholler>
I've tried 1.14 with .17 and 1.15 with .31
<aholler>
but as said, maybe I've missed a patch they've added to yocto
<jnettlet>
there shouldn't need to be any yocto specific patches.
<aholler>
not yocto-specific, but that seems the only place they post patches at all
<jnettlet>
in general on of the best tools to debug with is ldd. Make sure all the GL binarys are linking to the proper Vivante provided libGL.so
<aholler>
that problem was just with .31 were I've had a stale lib from .17
<jnettlet>
well .31 is a entirely different that is the v5 drivers which aren't binary compatible with v4 userspace. Have they released Xorg drivers for v5 yet?
<jnettlet>
specifically vivante_dri.so and libGL.so
<aholler>
yes
<aholler>
but the dri doesn't load at all, because of linking problems
<jnettlet>
well that is why it doesn't work then.
<jnettlet>
this has been a problem with Vivante for the last 5 years
<jnettlet>
I sent them patches to fix their linking problems that long ago and they refuse to accept anything
<jnettlet>
etnaviv is coming along well, so we won't have to worry about it soon
<aholler>
it's really about time that they get some pressure from intel ;)
<jnettlet>
ah, vivante doesn't care. There GPU's are the fastest in the world. But their size and power consumption is really well ahead of most the industry
<aholler>
it doesn't help much if they are fast if nobody is able to use them without spending years for setup ;)
<jnettlet>
although we are now running into the problem on the iMX6 that we have optimized the software so much that we are able to drive the SOC to hit thermal limits much faster and easier
<jnettlet>
but the setup problem is the same problem nvidia has had for years. full OSS drivers will make that much easier
<aholler>
I've stopped early trying to use play with self-made stuff on my nvidia tablet.
<aholler>
s/use//
<jnettlet>
yes, but even nvidia is coming around. They provided the patchs to get their K1 boards working with nouveau
<jnettlet>
we have a good community built around etnaviv right now. I think you will be impressed in what happens over the next 2-6 months
<aholler>
would be nice. I attempt to replace my current workhorse (a dockstar, kirkwood armv5 with 128mb) in that timeframe ;)
<aholler>
would already be able to do, but I first still want to play with the graphics. Even if I wont't use them afterwards ;)
<aholler>
but many times I think I would be happy to just use the i.mx6 as desktop. I live most of the time in vim anyways ;)
<aholler>
kde is just for konsole and alt-f2 gg: ;)
<aholler>
that should work with the i.mx6 too
<aholler>
(it already does, but sw-rendering doesn't make me happy)
<aholler>
anyway, thanks for all the tipps ;)
<aholler>
will test kwin_gles when the xserver starts ;)
<aholler>
hmm, I just remembered I compile the dri myself. So I can fix the linking problem myself.
<aholler>
looks like 3.10.17 has to wait a bit more ;)
<aholler>
just having to speak about switching from 3.10.17 to 3.10.31 is ridiculous ;)
<aholler>
and intend to end beta-test for 3.10.31 when 3.20 will be appear ;)
<aholler>
no containers for imx-users with the freescale kernel :/
<aholler>
nor nxtables and all the other nice stuff :(
aholler has quit [Ping timeout: 250 seconds]
paulk-collins has quit [Remote host closed the connection]
aholler has joined #imx6-dev
victhor_ has joined #imx6-dev
victhor_ is now known as victhor
bfederau_ has quit [Remote host closed the connection]
bfederau has joined #imx6-dev
jnettlet has quit [Ping timeout: 252 seconds]
victhor has quit [Read error: Connection reset by peer]