<ndufresne>
paulk-leonov, so now I got one stream working (it's H264 with P frames only and dpb size of 1 ;-D)
lurchi_ has joined #linux-sunxi
lurchi_ is now known as lurchi__
<ndufresne>
paulk-leonov, so is there known issues with H264 ? the other formats seems to work better
gnufan_home has quit [Quit: Leaving.]
<ndufresne>
paulk-leonov, I just tested with vaapi through ffmpeg, and same result, so I suspect something broken, maybe a regression of the index -> timestamp change, I'll try and update everything another day
_whitelogger has joined #linux-sunxi
mic-e[m] has quit [Ping timeout: 252 seconds]
t4h4[m] has quit [Ping timeout: 252 seconds]
JuniorJPDJ has quit [Ping timeout: 250 seconds]
thefloweringash has quit [Ping timeout: 250 seconds]
freddor has quit [Ping timeout: 252 seconds]
davidebeatrici has quit [Ping timeout: 252 seconds]
fevv8[m] has quit [Ping timeout: 252 seconds]
Jeremy_Rand_Talo has quit [Ping timeout: 252 seconds]
romainmahoux[m] has quit [Ping timeout: 252 seconds]
z3ntu has quit [Ping timeout: 250 seconds]
YannSoubeyrand[m has quit [Ping timeout: 252 seconds]
oliv3r[m] has quit [Ping timeout: 264 seconds]
MartijnBraam has quit [Ping timeout: 252 seconds]
psydroid has quit [Ping timeout: 252 seconds]
clementp[m] has quit [Ping timeout: 276 seconds]
solderfumes has quit [Ping timeout: 276 seconds]
EmilKarlson has quit [Ping timeout: 276 seconds]
k40s[m] has quit [Ping timeout: 248 seconds]
jernej has quit [Read error: Connection reset by peer]
jernej has joined #linux-sunxi
shfil has quit [Ping timeout: 264 seconds]
Mandragor_ has joined #linux-sunxi
shfil has joined #linux-sunxi
Mandragor has quit [*.net *.split]
nuuuciano__ has quit [*.net *.split]
s_frit has quit [*.net *.split]
sunshavi has quit [*.net *.split]
willmore has quit [*.net *.split]
ElBarto has quit [*.net *.split]
tlwoerner has quit [*.net *.split]
Wizzup has quit [*.net *.split]
buZz has quit [*.net *.split]
itdnhr_ has quit [*.net *.split]
duracrisis has quit [*.net *.split]
xeniter has quit [*.net *.split]
xeniter has joined #linux-sunxi
shfil has quit [Quit: Connection closed for inactivity]
kaspter has joined #linux-sunxi
romainmahoux[m] has joined #linux-sunxi
mic-e[m] has joined #linux-sunxi
oliv3r[m] has joined #linux-sunxi
k40s[m] has joined #linux-sunxi
JuniorJPDJ has joined #linux-sunxi
thefloweringash has joined #linux-sunxi
t4h4[m] has joined #linux-sunxi
Jeremy_Rand_Talo has joined #linux-sunxi
davidebeatrici has joined #linux-sunxi
fevv8[m] has joined #linux-sunxi
freddor has joined #linux-sunxi
MartijnBraam has joined #linux-sunxi
z3ntu has joined #linux-sunxi
YannSoubeyrand[m has joined #linux-sunxi
psydroid has joined #linux-sunxi
buZz has joined #linux-sunxi
buZz has quit [Changing host]
buZz has joined #linux-sunxi
clementp[m] has joined #linux-sunxi
solderfumes has joined #linux-sunxi
EmilKarlson has joined #linux-sunxi
tllim has quit [Read error: Connection reset by peer]
duracrisis has joined #linux-sunxi
Wizzup has joined #linux-sunxi
ElBarto has joined #linux-sunxi
sunshavi has joined #linux-sunxi
tlwoerner has joined #linux-sunxi
nuuuciano__ has joined #linux-sunxi
itdnhr_ has joined #linux-sunxi
willmore has joined #linux-sunxi
s_frit has joined #linux-sunxi
s_frit has quit [Remote host closed the connection]
s_frit has joined #linux-sunxi
victhor has quit [Ping timeout: 257 seconds]
tllim has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
lurchi_ has joined #linux-sunxi
lurchi__ has quit [Ping timeout: 268 seconds]
loki__ has joined #linux-sunxi
cnxsoft has joined #linux-sunxi
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 - http://znc.in]
MoeIcenowy has joined #linux-sunxi
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 - http://znc.in]
MoeIcenowy has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
wigyori has quit [Ping timeout: 255 seconds]
wigyori has joined #linux-sunxi
tllim has quit [Read error: Connection reset by peer]
RichardG867 has quit [Quit: Keyboard not found, press F1 to continue]
RichardG867 has joined #linux-sunxi
cmeerw has joined #linux-sunxi
reinforce has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
pmpp has quit [Disconnected by services]
pmpp_ has joined #linux-sunxi
bonbons has quit [Quit: Leaving]
_whitelogger has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
kaspter has quit [Quit: kaspter]
nuuuciano__ has quit [Ping timeout: 245 seconds]
parazyd has quit [Read error: error:1408F10B:SSL routines:ssl3_get_record:wrong version number]
parazyd has joined #linux-sunxi
loki__ has quit [Remote host closed the connection]
loki__ has joined #linux-sunxi
loki__ has quit [Ping timeout: 246 seconds]
kaspter has joined #linux-sunxi
dddddd has quit [Remote host closed the connection]
vagrantc has quit [Quit: leaving]
sunshavi has quit [Remote host closed the connection]
_whitelogger has joined #linux-sunxi
imcsk8 has quit [Ping timeout: 246 seconds]
imcsk8 has joined #linux-sunxi
Andy-D has joined #linux-sunxi
Mangy_Dog has joined #linux-sunxi
lurchi_ is now known as lurchi__
s_frit has quit [Remote host closed the connection]
s_frit has joined #linux-sunxi
lurchi__ is now known as lurchi_
JohnDoe_71Rus has quit [Ping timeout: 246 seconds]
JohnDoe_71Rus has joined #linux-sunxi
tilpner has quit [Quit: WeeChat 2.4]
victhor has joined #linux-sunxi
smaeul has quit [Ping timeout: 252 seconds]
smaeul has joined #linux-sunxi
mpmc has quit [Ping timeout: 268 seconds]
lurchi_ is now known as lurchi__
<Mangy_Dog>
Need a little confirmation as im not 100% sure, OTG... If i want to connect my PI to the computer, and have the computer read the pi as a data storage device, do i tie the ID pin to gnd or float it, on the pi side?
tilpner has joined #linux-sunxi
<aalm>
don't you have a connector for it?
<aalm>
i think cables/adapters connected to your pi via connector on its board should do the right thing for you
<aalm>
==floating, unless i'm wrong:]
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
tilpner has quit [Ping timeout: 258 seconds]
<Mangy_Dog>
nono its not that at all
<Mangy_Dog>
im fitting my orange pi into a interface pcb
lurchi__ is now known as lurchi_
<Mangy_Dog>
that pcb is going to have a usb socket in it to for charging the device
<Mangy_Dog>
i want to have that double as a data line to the storage on the device
<Mangy_Dog>
so passing through te data pins to the otg plug on the pi
<Mangy_Dog>
i need to know if i need to pass through or pull up or pull down the id pin
<Mangy_Dog>
though im not thinking just pass it through
cnxsoft has quit [Remote host closed the connection]
gnufan_home has quit [Ping timeout: 245 seconds]
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 - http://znc.in]
MoeIcenowy has joined #linux-sunxi
lurchi__ is now known as lurchi_
AneoX_ has quit [Read error: Connection reset by peer]
AneoX has joined #linux-sunxi
lurchi_ is now known as lurchi__
mpmc has quit [Ping timeout: 246 seconds]
jstein_ has joined #linux-sunxi
jstein_ is now known as jstein
mpmc has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
tllim has joined #linux-sunxi
<willmore>
Mangy_Dog, it will vary by device, but most SBCs that do OTG will either *assume* device mode or will use a GPIO to sense their role.
<willmore>
You'll have to see what your hardware does. Note: if it uses an A port for the OTG, there's no sense possible as the fifth pin is not present on the connector.
<Mangy_Dog>
the pi ignores the ID pin>
<Mangy_Dog>
?
<willmore>
It might
<Mangy_Dog>
hmm ok
<willmore>
Depends on how that particular SBC is wired.
<Mangy_Dog>
well im just going to pass the id to the pi board anyway so i cant fekk that on eup
<willmore>
If you're lucky, it's at least hooked to a GPIO
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 - http://znc.in]
MoeIcenowy has joined #linux-sunxi
<willmore>
Might as well pass the signal along. You might also break it off to a separate header and hook that to a GPIO of your choice.
<willmore>
You could then use that in a script to change the OTG mode.
nuuuciano__ has joined #linux-sunxi
lurchi__ is now known as lurchi_
nuuuciano__ has quit [Ping timeout: 246 seconds]
nuuuciano__ has joined #linux-sunxi
foxx_ has joined #linux-sunxi
nuuuciano_ has joined #linux-sunxi
curlybracket has quit [Ping timeout: 258 seconds]
nuuuciano has joined #linux-sunxi
nuuuciano__ has quit [Ping timeout: 246 seconds]
nuuuciano_ has quit [Ping timeout: 246 seconds]
nuuuciano_ has joined #linux-sunxi
Putti has joined #linux-sunxi
nuuuciano has quit [Ping timeout: 244 seconds]
lurchi_ is now known as lurchi__
lurchi__ is now known as lurchi_
nuuuciano_ has quit [Ping timeout: 258 seconds]
return0e has quit [Remote host closed the connection]
nuuuciano_ has joined #linux-sunxi
<dddddd>
anarsoul, MoeIcenowy I'm testing 5.1.1 in the Olimex TERES-I and the dts still lacks backlight and &pwm nodes <https://lkml.org/lkml/2018/3/15/666>. I seems they weren't merged back then (the reason escapes me, but there's some info in that thread). I just added them and now I can use /sys/class/backlight/* (empty without those nodes).
nuuuciano_ has quit [Ping timeout: 258 seconds]
<anarsoul>
submit a patch?
<dddddd>
anarsoul, I'm not sure if that's the right way to configure them, just tested. Also, I'm not that familiar with the procedure. Is the pwm situation clear now?
<anarsoul>
dddddd: use pinebook dts as reference
<dddddd>
anarsoul, thanks. I'll try to mimic its format. I'll test right now.
lurchi_ is now known as lurchi__
<dddddd>
anarsoul, perfect; &pwm only needs the status line. About the backlight one... Harald's patch (linked above) does not include power-supply (pinebook dts has it). Do I need to add one?
wwilly has joined #linux-sunxi
<anarsoul>
dddddd: check schematics
<anarsoul>
if you have gpio to enable backlight power then you need regulator-fixed
<dddddd>
anarsoul, I mean... it's working without power-supply. I'm wondering if it's required. I'll look at the schematics, with some luck I'll understand something... hehe
<anarsoul>
dddddd: u-boot probably left it on for you
<dddddd>
anarsoul, when trying modern u-boot I'm using mainline ATF. I guess I can try recent ATF with older/working u-boot.
<paulk-leonov>
ndufresne, so that's with cedrus?
<paulk-leonov>
ndufresne, there's a chance we only updated v4l2-request-test for the last round
<paulk-leonov>
did you get that working already?
<paulk-leonov>
it's good to have as a reference
<dddddd>
anarsoul, like... 2018-01 works for me, that's what I'm refering too.
<anarsoul>
dddddd: that's pretty old
<dddddd>
anarsoul, I'll re-check if I'm using patched u-boot (other than dts). As I said, the just merged one doesn't work at all (not even serial output)
<anarsoul>
well, try contacting MoeIcenowy, she has Teres and can confirm whether it works for her or not
lurchi__ is now known as lurchi_
<dddddd>
anarsoul, I will. Thank you very much!
mpmc has quit [Ping timeout: 246 seconds]
<Mangy_Dog>
the raw audio LR out of orange pi 13 pin header.... Is that signal strong enough for headphones? or does a headphone port need an amp?
<Mangy_Dog>
im amping to speakers, but if i can just use a switching jack to take the audio feed off the pi direct that would be useful
tl_lim has joined #linux-sunxi
tllim has quit [Ping timeout: 252 seconds]
<dddddd>
aalm, yes... thank you. I even named her in a previous message just in case she was reading.
<MoeIcenowy>
in fact for LInux DT I want to wait for anx6245
<MoeIcenowy>
6345 *
CSoto_ has joined #linux-sunxi
lurchi_ has quit [Quit: Konversation terminated!]
lurchi_ has joined #linux-sunxi
<dddddd>
Hello MoeIcenowy! Do you know any reason why u-boot 504bf790da might not work for teres-i? (no serial output here)
<dddddd>
(I guess it works for you)
clemens3 has joined #linux-sunxi
clemens3 has quit [Client Quit]
<MoeIcenowy>
lemme check which commit it is
clemens3 has joined #linux-sunxi
mpmc has joined #linux-sunxi
<MoeIcenowy>
dddddd: not checked yet
<MoeIcenowy>
should work
<dddddd>
MoeIcenowy, OK... thanks. Let me know the outcome if you try it. I'll keep you posted too if I find the cause.
lkcl has quit [Read error: Connection reset by peer]
jstefanop has quit [Remote host closed the connection]
jstefanop has joined #linux-sunxi
BenG83 has joined #linux-sunxi
lkcl has joined #linux-sunxi
nuuuciano has joined #linux-sunxi
netlynx has quit [Quit: Ex-Chat]
lurchi_ is now known as lurchi__
lurchi__ is now known as lurchi_
kaspter has quit [Ping timeout: 258 seconds]
kaspter has joined #linux-sunxi
cmeerw has quit [Ping timeout: 252 seconds]
Mangy_Dog has quit [Ping timeout: 245 seconds]
<ndufresne>
paulk-leonov, ok, I'll give a look at the most recent changes on v4l2-request-test so see what changed, what's I was mostly worried is if I simply picked the wrong kernel, I picked the one on bootlin github, the one called next
lurchi_ is now known as lurchi__
<paulk-leonov>
ndufresne, mhh I think the latest one we released is that latest release tag
<paulk-leonov>
I should really move my git to a better place to put that there
<ndufresne>
well, if you feel like pointing me to one of your known to work kernel branch, I'll take
<ndufresne>
paulk-leonov, my next step if I can't figure-out, is to build Kwiboo's ffmpeg, and see if that works, from there I should be able to find what going wrong
<ndufresne>
it just made a little worried that the index -> timestamp move might have came with regressions
<ndufresne>
when I look at it now, it seems a tad more fragile (but less likely to endup reusing random images)
lurchi__ is now known as lurchi_
<paulk-leonov>
well the way I implemented it initially was really a quick dirty hack
<paulk-leonov>
but yeah that libva would need some serious rework
<paulk-leonov>
I'd be more than happy to kill it ASAP
<paulk-leonov>
ndufresne, btw about h264 DPBs, it's still unclear to me what data we should be passing to the kernel
<ndufresne>
I don't know, I know you don't like it, but I don't currently have the time to write native support for gst
<paulk-leonov>
so apparently there's a fashion of the DPB where we have all entries including unused ones and one where it's only the entries for the current frame
<paulk-leonov>
yeah I think ffmpeg is the way to go for now
<paulk-leonov>
well anyway, I'll see if I can get some time to update the userspace and kernel tree tomorrow
<ndufresne>
well, the implementation is not that great, and not a great fit for gst
<ndufresne>
ffmpeg does not have any memory referencing model
<paulk-leonov>
mhh I see
vagrantc has joined #linux-sunxi
<paulk-leonov>
ah?
<paulk-leonov>
like, you can't reference buffers?
<paulk-leonov>
I thought there were helpers for that
<ndufresne>
for us in GStreamer model it's been a pain to deal with, and for zero copy path, it's really complicated
<ndufresne>
anyway, it's not that relevant
<ndufresne>
I think there is some benifit in having that VA driver around and maintained in the very short term
<paulk-leonov>
okay then
foxx_ has quit [Ping timeout: 246 seconds]
<ndufresne>
it's takes 5s to build, and I can use the distro ffmpeg or gstreamer to use it (after the fix I just sent)
<paulk-leonov>
heh fair enough
<paulk-leonov>
ndufresne, good to know about rk3288 being a hantro btw
<ndufresne>
some of the mistake there are, decoding should start before end_frame, end_frame should not block
<paulk-leonov>
I just emailed the Purism guy to let him know what's up
<paulk-leonov>
ndufresne, yeah of course
<paulk-leonov>
but that's just the overall design issue
<paulk-leonov>
since we're not operating per-slice yet
<ndufresne>
paulk-leonov, yes, I received a gift board IMX8M and downloaded the manual and was like, eh, this is the same registers I've been reading over and over for the last week
<paulk-leonov>
hehe :)
<ndufresne>
anyway, pH5 will be starting soon on IMX8M support, so I guess he'll work closely with bbrezillon (for him it will be such a time saver)
<ndufresne>
the RK thing is basically 3 drivers, because there is different chips
<ndufresne>
but it's tricky, since you need to do time-division of them
<ndufresne>
the register region overlaps ;-D
wwilly has quit [Ping timeout: 258 seconds]
<ndufresne>
paulk-leonov, btw, right now bbrezillon is using Kwiboo's ffmpeg for testing, and he's trying to get passed the very basic now, last Friday we had the impression that the controls and the slices was not being matches properly
<ndufresne>
it's all new code
<paulk-leonov>
nice, great to hear
<paulk-leonov>
yeah I know rockchip's a mess with so many different blocks
<Kwiboo>
ndufresne, to my knowlage it should be working correctly, at least when tested on cedrus the decoding seems to be working, but I also know that jernej reverted a timestamp_copied kernel patch to fix one issue
<paulk-leonov>
I think I have an exotic rk3229 board with an unusual setup too
<ndufresne>
Kwiboo, ok, is that in LibreELEC ? my goal is to reach sane state, from there I should be in better place to help ;-D
<ndufresne>
rk3229, never heard of that one ;-D but I'm still quite new to RK thingy
<ndufresne>
we didn't have manual for RK3288, but found that there's another one, called PX3 which have the same codec, it's a bit like amlogic, so many variants of the same thing
popolon has joined #linux-sunxi
<Kwiboo>
ndufresne, yep in LibreELEC, btw to ease testing you can add avctx->skip_frame = AVDISCARD_NONINTRA; to h264 init function to filter out all but I-frames (or use any of the other discard options)
<Kwiboo>
start/end_decode will not be called for other frame types, at least helped me when I tested on cedrus
<Kwiboo>
not sure what the issue was, but one file that jernej had a regression with that commit, guessing there may be a reference frame missing and the first one returned without the copied_timestamp set produced better output
gnufan_home has joined #linux-sunxi
<jernej>
without the revert, I got a lot of green squares
<jernej>
but video was interlaced, so I think it was reference issue
<jernej>
again with referencing itself (top/bottom)
<jernej>
@Kwiboo how does ffmpeg deal with interlaced h264? single or double end_frame call?
<Kwiboo>
jernej: was it referencing itself or another frame?, I am wondering if without copied_timestamp it happen to find a buffer that happens to have similar content instead of a empty buffer
<Kwiboo>
jernej, to my knowlage there will be two start/end calls, one for top/bottom field
<jernej>
I didn't do any debugging, but I think it references itself
<jernej>
yeah, I'm wondering if timestamps are broken in ffmpeg
<jernej>
I have few H264 videos which doesn't work for no apparent reason
<jernej>
and for one video wrong references would explain corruption
<Kwiboo>
was there code added that requires timestamp > 0 or similar ?
<Kwiboo>
since we use buffer index first one is 0
<Kwiboo>
I can do some debugging tomorrow, I have not touched the ffmpeg code in a while so will be fun :-)
lurchi__ is now known as lurchi_
<ndufresne>
ok, thanks, cost nothing for me to try that rever
<ndufresne>
Kwiboo, jernej: it's up to the encoder to combine or not the fields btw
<ndufresne>
so you may have 1 slice per field, a 1 slice with two fields
<Kwiboo>
jernej, thanks, with that video you can see the green blocks vs not depending on the copied_timestamp commit?
<jernej>
no, that's another one
<jernej>
just a sec
<Kwiboo>
thanks, I will check how the timestamps flows in/out of ffmpeg, not sure I have fully traced it since the swtich to timestamps, and lots have happened on the kernel side since then
<jernej>
anyway, issue can be noticed just first second or two
lurchi_ is now known as lurchi__
<jernej>
after that it's ok
lurchi__ is now known as lurchi_
valkyr1e has quit [Quit: Bye.]
pmpp_ has quit [Ping timeout: 246 seconds]
valkyr1e has joined #linux-sunxi
lurchi_ is now known as lurchi__
pmpp has joined #linux-sunxi
clemens3 has quit [Quit: WeeChat 1.6]
pmpp has quit [Ping timeout: 246 seconds]
deesix has joined #linux-sunxi
pmpp has joined #linux-sunxi
<ndufresne>
So even with that patch, playback isn't right for most stream with the VA driver, this will need to be debugged, I'll build ffmpeg "native" backend now to confirm
<ndufresne>
paulk-leonov, about the reference list, I don't think it really matter if it's a complete DPB, or just the one being referenced for this frame, we can leave that to usespace
<ndufresne>
now the naming DPB might be confusing
<ndufresne>
in the case of VAAPI it was designed with a client/server model (X11) where all the data has to be uploaded to the server, so they try and make everything as small as possible
BenG83 has quit [Quit: Leaving]
<ndufresne>
the other thing I have seen, is that the created list subset is often sorted to speedup lookup with building the ref lists, I'm not sure how much of this is significant
<ndufresne>
paulk-leonov, what I think is that for the accelerator it's just an array, if there is no reference to some entry in l0/l1 it will just be ignored
xcko has joined #linux-sunxi
reinforce has quit [Quit: Leaving.]
Kwiboo has quit [Ping timeout: 268 seconds]
pmpp has quit [Disconnected by services]
pmpp_ has joined #linux-sunxi
jo0nas has joined #linux-sunxi
jstein has quit [Quit: quit]
pmpp has joined #linux-sunxi
pmpp_ has quit [Ping timeout: 255 seconds]
lurchi__ is now known as lurchi_
popolon has quit [Quit: WeeChat 2.4]
lurchi_ has quit [Read error: Connection reset by peer]