ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development - Did you try looking at our wiki? https://linux-sunxi.org - Don't ask to ask. Just ask and wait for an answer! - This channel is logged at https://oftc.irclog.whitequark.org/linux-sunxi
sunshavi_ has quit [Ping timeout: 480 seconds]
Danct12 has quit [Read error: Connection reset by peer]
pgwipeout[m] has joined #linux-sunxi
Tooniis[m] has joined #linux-sunxi
Arthur[m]123 has joined #linux-sunxi
cmeerw[m] has joined #linux-sunxi
ftg has quit [Ping timeout: 480 seconds]
cnxsoft has joined #linux-sunxi
JuniorJPDJ has joined #linux-sunxi
moteen has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
vagrantc has joined #linux-sunxi
hexdump01 has joined #linux-sunxi
hexdump0815 has quit [Ping timeout: 480 seconds]
Danct12 has joined #linux-sunxi
vagrantc has quit [Quit: leaving]
aggi has quit [Remote host closed the connection]
aggi has joined #linux-sunxi
psydroid[m] has joined #linux-sunxi
moteen has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
rajkosto has quit [Read error: Connection reset by peer]
chuang[m] has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
moteen has joined #linux-sunxi
<swiftgeek> Is there any h264enc + modern kernel (with extra module) known to work on A20?
<swiftgeek> the furthest I got so far was with https://github.com/uboborov/sunxi-cedar-mainline
z3ntu has joined #linux-sunxi
<swiftgeek> but I get a hang like this https://bpa.st/6QFQ
t4h4[m] has joined #linux-sunxi
<swiftgeek> Loading the module looks like this in dmesg https://bpa.st/Y6IA and h264enc starts with this https://bpa.st/IPZA
<swiftgeek> it doesn't crash with no input though lol
<swiftgeek> *hang
Danct12 has quit [Remote host closed the connection]
Danct12 has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
moteen has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
moteen has joined #linux-sunxi
moteen has quit [Ping timeout: 480 seconds]
<swiftgeek> memtest doesn't appear to have any issues on this board
<swiftgeek> * memtester
bauen1 has quit [Ping timeout: 480 seconds]
Danct12 has quit [Remote host closed the connection]
Danct12 has joined #linux-sunxi
<swiftgeek> ok after using makefile from there in https://github.com/libv/cedarx_h264_encoder/tree/master/kernel , I got it to compile and load, https://bpa.st/IZWQ
<libv> heh
<libv> it was working on the Q4 2020 kernel
<libv> even has working thumbnails, which allwinner never tested
<libv> and the tree was started with actual archeology, which no-one seems to care for these days
<libv> so the cedrus people actually have their real commits preserved
<libv> mostly jemk's code though
<libv> people like uboborov and all the other copies mostly dropped history and added nothing new
<swiftgeek> welp until I make some silly systemrdl tools to aid me, there is no way I can make any sense of it :D
<libv> i should resume work on this soon, now that life is slowly turning back to normal
<swiftgeek> but I saw in git commits that it focuses on A20 entirely, and I will soon lose access to that
<libv> hot topic now is getting the wiki migrated though
<swiftgeek> and H3 is still waiting for replacement balls
<libv> swiftgeek: this code is for the fosdem video hw project
<libv> which is a20
<swiftgeek> which leaves me with just few A13 boards :D
<libv> and a20 also has G2D and the more useful disp engine (which can handle planar RGB)
<libv> a13 has the same bits
<swiftgeek> one thing i noticed as difference between your repo and the other one was SRAM being released in yours
<swiftgeek> and in other it doesn't seem to ever happen
<swiftgeek> so got a lot of strange behaviour
<libv> it's been 20 months since i last worked on this
<libv> it should not be released?
<swiftgeek> but the other repo has ffmpeg with code from h264enc integrated, making it easier to use ;D
<swiftgeek> or was it a patchset
<libv> these resource bits were taken from the free-electrons kickstarter-ish work
<libv> the rest is extremely reworked from jemks work
<libv> the ffmpeg stuff was pretty nastily done
<swiftgeek> huh
<swiftgeek> no crashes
gsz has joined #linux-sunxi
<libv> ?
<swiftgeek> with code from your repo i get not crashes and some output is happening
<libv> was working on 5.9.0
<swiftgeek> welp the output happens to be 0x00 xD
<swiftgeek> but at least it's not crashing
<libv> there is nothing in the dts that pertains to the encoder
<swiftgeek> a test input known to work would be nice to have (i guess captured output of ffmpeg before h264enc would work fine)
<libv> this is i think what is still valid in the readme
<libv> the rest in the readme is not relevant at all
<swiftgeek> welp right now I have no way to tell whether I'm feeding h264enc the right data
<swiftgeek> unless everything works fine
<swiftgeek> so with known good input (i guess few raw nv12 frames?) in git repo or sth, there would be less to worry about
<swiftgeek> and even if I try to put ffmpeg output somewhere, ffprobe doesn't recognize nv12 anywhere in it
ftg has joined #linux-sunxi
<swiftgeek> I mean something like ./h264enc ~/infile_640x360.nv12 640 360 /tmp/test.x264 with infile_640x360.nv12 being in repo
bauen1 has joined #linux-sunxi
<swiftgeek> yeah
<swiftgeek> looked at other side, and nv12 is definitely not right xD
<swiftgeek> 0x10 there
<libv> the bit of big buck bunny as nv12 that i use is 1.5GB
ftg has quit [Ping timeout: 480 seconds]
<libv> for 1k frames
<libv> 1280*720*1.5 / 1024 = 1350MB
<swiftgeek> yeah but just some few frames for such test case would be enough
<libv> the last bit of work i did was thumbnailing
<libv> which is needed for the fosdem video box lcd
<libv> to vastly decrease memory bandwidth usage
<libv> upcoming work will be adding full sysfs config and io, and then add dmabuf input
<libv> and then i can tie it in with the rest of the fosdem video pipeline
<swiftgeek> ffmpeg has -frames so it should be possible to seek somewhere in the middle and output few nv12 frames precisely
<libv> swiftgeek: i just took big buck bunny and used that documented ffmpeg command to generate the file
<libv> there was nothing to it
<libv> and i too initially went through that emotion that you have now
<swiftgeek> yeah when it works there is nothing to it xD
<swiftgeek> but few frames in repo would save many emotions
<libv> again, this is not meant for users
<libv> this is a step along the long road to build the fosdem video box
<swiftgeek> well making it easier to reproduce working is still a good thing
<libv> which is capturing an hdmi signal, then displaying on hdmi-out while turning some frames into nv12 using G2D, then encoding it to h.264, and sending the result over the network while displaying some 4x scaled down frames on the status lcd
<swiftgeek> and if output is deterministic, then it's also a simple pass/fail
<libv> patches are welcome
<swiftgeek> welp I need to reproduce it first ;D
aggi has quit [Remote host closed the connection]
aggi has joined #linux-sunxi
<swiftgeek> and I don't even know whether it would work with just one frame or whatever, and while I can grab big buck bunny, "bigbuckbunny.mpg" in google resolves to just your repo :D
<libv> there is little i can do here.
<libv> *sigh* let me put this on our dl server.
<libv> it will take 10mins
<libv> at leastt
<swiftgeek> tried creating nv12 frame with webcam until it looked fancy in xxd, still getting 0x00 on output side xD
<libv> the last time i used this file was january 2021
<libv> the last time i used this code was january 2021
<libv> y,,v
<libv> ymmv
<swiftgeek> lol tried passing ffmpeg nv12 output i created myself back to ffplay and it doesn't like it
<swiftgeek> and yours works with ffplay -f rawvideo -video_size 1280x720 -pixel_format nv12 ./big_buck_bunny.1280x720.1024.nv12
evgeny_boger1 has joined #linux-sunxi
evgeny_boger has quit [Ping timeout: 480 seconds]
evgeny_boger1 has quit [Remote host closed the connection]
evgeny_boger has joined #linux-sunxi
evgeny_boger has quit [Remote host closed the connection]
<libv> swiftgeek: did you manage to encode this?
<libv> md5sum of the result should be 29d81ffe24794281e4af0d7e51a74f84
<libv> last verified on 20210101
<libv> oldest file i have with the same md5sum is 20191110, shortly after the first cleanup/archeology
<libv> this md5sum will see some changes in future as the h264 parameters need filling out
JohnDoe_71Rus has quit []
Danct12 has quit [Remote host closed the connection]
<swiftgeek> > Frame 682: 35bytes
<swiftgeek> that doesn't look promising
<swiftgeek> yeah 0x00 just like before
<swiftgeek> I guess I need to try downgrading
pg12 has joined #linux-sunxi
<libv> swiftgeek: it should be largely independent of the actual kernel
<libv> perhaps there is a difference in dts that is not caught by the resource code?
<swiftgeek> then i would expect errors / module doing nothing
<swiftgeek> which is what happened initially because i had DT modified for the other module
<swiftgeek> oh LOL
<swiftgeek> i had a hunch now that it might be opposite to what i thought
<swiftgeek> not only it wants cedrus module, but it wants it loaded before cedar module
<swiftgeek> got output!
<swiftgeek> > Stream #0:0: Video: h264 (Main), yuv420p(progressive), 1280x720, 25 fps, 25 tbr, 1200k tbn, 50 tbc
<swiftgeek> and ffprobe recognizes it
<swiftgeek> that's 5.17
<swiftgeek> and it plays properly with ffplay on a lappy
<swiftgeek> 40 seconds
<libv> so it needs cedrus loaded... hrm
<libv> i guess i never tested that
<libv> but why...
<libv> some powermanagement stuff perhaps
<libv> i could have sworn i had it blacklisted
arnd__ has quit []
<libv> on the other hand, the hardware was incredibly stable, both for bringing up g2d and this
arnd has joined #linux-sunxi
<libv> so it had massive uptimes and it perhaps was not necessary to do more than ctrl-r cedrus and pressing enter
Danct12 has joined #linux-sunxi
<libv> i added an issue to github
<libv> because it will be a few weeks before i get to this
gamiee_ has quit [Remote host closed the connection]
gamiee has joined #linux-sunxi
<gamiee> test
<gamiee> libv: I checked your code, and I don't see your kernel module to configure DRAM, what is required. I guess cedrus does that.
<gamiee> so most likely, that's why it doesn't work without cedrus kernel module.
cnxsoft1 has joined #linux-sunxi
cnxsoft has quit [Ping timeout: 480 seconds]
pnill has joined #linux-sunxi
<libv> ok
<swiftgeek> is H3 or A13 easier to get working from this point? :D
<gamiee> swiftgeek: I remember that A20 have some differences (or erratas?) in Video Engine, so H3 might be easier
<libv> hrm, maxima seems gone now
<swiftgeek> well if A20 has erratas there i would expect same/similar errata set on A13, and then H3 being "proper"
<swiftgeek> huh that's odd
<swiftgeek> in earlier rev it says 1080p encoding, but later on 720p
<libv> gamiee: i am not seeing it
<gamiee> libv: as far as I remember, without this: https://github.com/gamelaster/pinecube/blob/main/software/terrenc/terrenc.c#L19 any read / write to VE result in deadlock
<libv> oh, ok, i would never have caught that
<libv> gamiee: thanks
<libv> is this for s3 only?
<gamiee> it was only tested on S3, and "terrenc" is deprecated (and replaced with recedar). But now when I'm looking at that, seems that you're already doing that here: https://github.com/libv/cedarx_h264_encoder/blob/master/kernel/cedar.c#L442
<gamiee> without cedar kernel module, any usage of encoder locks up ?
<libv> i do not have the hw setup, it's been 18 months since i touched this
<libv> swiftgeek should know more
<swiftgeek> without cedrus module i just get 0x00 (or cedrus loaded after cedar)
<swiftgeek> and I didn't try other repo stuff meant for primarily S3, together with cedrus yet (coz i commented out that section)
<swiftgeek> Though not sure if i can have video-engine@1c0e000 and video-codec@1c0e000 together, that sounds like a conflict
JohnDoe_71Rus has joined #linux-sunxi
<swiftgeek> and reg portion is different , not sure if i can extend it like that cedrus
<swiftgeek> I guess I should bother xunlong about boardview export in GenCAD format with artwork stripped, so I can revive H3 board without much stress
<swiftgeek> would be nice if they had IRC contact xD
<libv> good point, i should pry this ve_mode register apart, at least document this
<libv> but this is always written
<libv> so that cannot be the cause
<swiftgeek> would be nice to know whether they used allegro for layout, then i could provide instructions for that exactly
aggi has quit [Ping timeout: 480 seconds]
Daanct12 has joined #linux-sunxi
Danct12 has quit [Ping timeout: 480 seconds]
cnxsoft1 has quit []
aggi has joined #linux-sunxi
bauen1 has quit [Ping timeout: 480 seconds]
moteen has joined #linux-sunxi
buZz has joined #linux-sunxi
vagrantc has joined #linux-sunxi
<jernej> wiki seems to be down
Daaanct12 has joined #linux-sunxi
ftg has joined #linux-sunxi
Daanct12 has quit [Ping timeout: 480 seconds]
bauen1 has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
Daanct12 has joined #linux-sunxi
moteen has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
Daaanct12 has quit [Ping timeout: 480 seconds]
<libv> jernej: indeed, machine is called maxima
<libv> 15:14 < libv> hrm, maxima seems gone now
<jernej> ah, I didn't know
<libv> nothing i can do about it atm, mnemoc needs to get involved there
Guest6561 is now known as palmer
bauen1 has quit [Ping timeout: 480 seconds]
moteen has joined #linux-sunxi
moteen has quit [Read error: Connection reset by peer]
moteen has joined #linux-sunxi
moteen has quit [Remote host closed the connection]
bauen1 has joined #linux-sunxi
evgeny_boger has joined #linux-sunxi
moteen has joined #linux-sunxi
moteen has quit [Ping timeout: 480 seconds]
Danct12 has joined #linux-sunxi
Daanct12 has quit [Ping timeout: 480 seconds]
JohnDoe_71Rus has quit []
hlauer has joined #linux-sunxi
moteen has joined #linux-sunxi
Luke-Jr has quit [Ping timeout: 480 seconds]
moteen has quit [Ping timeout: 480 seconds]
gsz has quit [Quit: leaving]
aggi has quit [Remote host closed the connection]
aggi has joined #linux-sunxi
sunshavi has joined #linux-sunxi
Luke-Jr has joined #linux-sunxi
ftg^ has joined #linux-sunxi
moteen has joined #linux-sunxi
HERIC has joined #linux-sunxi
gamiee_ has joined #linux-sunxi
pg12_ has joined #linux-sunxi
ftg^ has quit [charon.oftc.net coulomb.oftc.net]
hlauer has quit [charon.oftc.net coulomb.oftc.net]
ftg has quit [charon.oftc.net coulomb.oftc.net]
gamiee has quit [charon.oftc.net coulomb.oftc.net]
pg12 has quit [charon.oftc.net coulomb.oftc.net]
z3ntu has quit [charon.oftc.net coulomb.oftc.net]
psydroid[m] has quit [charon.oftc.net coulomb.oftc.net]
Arthur[m]123 has quit [charon.oftc.net coulomb.oftc.net]
Tooniis[m] has quit [charon.oftc.net coulomb.oftc.net]
hlauer has joined #linux-sunxi
karlp has quit [Ping timeout: 480 seconds]
moteen has quit [Ping timeout: 480 seconds]
bauen1 has quit [Ping timeout: 480 seconds]
hlauer has quit [Read error: Connection reset by peer]
hlauer has joined #linux-sunxi
bauen1 has joined #linux-sunxi
karlp has joined #linux-sunxi
hlauer has quit [Ping timeout: 480 seconds]
psydroid[m] has joined #linux-sunxi
z3ntu has joined #linux-sunxi
mnemoc has joined #linux-sunxi
<swiftgeek> looks like libv cedar module kicked in with combined DT section like this https://bpa.st/4OPQ , and so does cedar_ve from the other repo. but cedar_ve/cedar and sunxi-cedrus don't work together - only first one that is loaded works
<swiftgeek> so since it applies to both maybe I should strip reg stuff and try again
<libv> swiftgeek: there are no changes to the dtses for the fosdem device that relate to cedarx
<libv> for me
<swiftgeek> lol can't reproduce what i had just few hours ago xD
<swiftgeek> oh nvm, i just didn't read properly what happened
<swiftgeek> so load sunxi-cedrus, unload, load cedar(ve) ?
<swiftgeek> yeah so loading cedar directly after boot gives me constant"35 bytes"/0x00, loading sunxi_cedrus after that does nothing, but loading sunxi_cedrus, unloading it then loading cedar makes cedar+h264enc provide proper output
evgeny_boger has quit [Ping timeout: 480 seconds]
<swiftgeek> so cedar_ve fails no matter what
Tooniis[m] has joined #linux-sunxi
<swiftgeek> but at least I don't need to swap DT to test between them anymore
<swiftgeek> can load all 3 modules but not at the same time :D
aggi has quit [Quit: connection closed.]
<swiftgeek> it's a bit jumpy but still faster than playback time
aggi has joined #linux-sunxi