ChanServ changed the topic of #cedrus to: Cedrus development channel (http://linux-sunxi.org/Cedrus) | *only registered users can talk* | Logs at https://oftc.irclog.whitequark.org/cedrus/
jernej_ has joined #cedrus
jernej has quit [Remote host closed the connection]
<ndufresne> jernej_: I've resetup my Olimex A20-OLinuXino-LIME2-eMMC, to test tiled decoding with cedrus, and am still getting garbage out of it
<ndufresne> is cedrus working on A20 and similar ?
<ndufresne> decoded images are all green with a slight corrupted area down right
jernej_ is now known as jernej
<jernej> ndufresne: last time I checked cedrus on Lime2, it worked
<jernej> but that was about a year back
<jernej> and with LE/Kodi/ffmpeg stack
<jernej> ndufresne: here I have last build images for A20: http://jernej.libreelec.tv/images/a20/
<jernej> we drop A20 from LE, because it required a ton of work and nobody was prepared to do that
<ndufresne> tone of work, you mean there is tone of downstream kernel patches ?
<jernej> well, depends what you want
<jernej> for example, HDMI audio has only initial patch but with a lot of problems
<jernej> there is thing with primary DRM plane that it can't be smaller than screen resolution (annoying if player chose primary plane for video playback)
<jernej> patch for this exist, but it has not made in mainline
<jernej> yet
<jernej> anyway, my preferred platform is H3 or newer, much easier to work since peripherals are more capable
<jernej> there are problems with memory bandwith (not sure if A20 or sibling), but full HD is on upper limit
<jernej> also deinterlacer is somehow part of display pipeline - no chance to have anything useful implemented for this anytime soon
<jernej> maybe via DRM properties?
<jernej> but some users would still like to see A20 as viable player and were very disappointed when I dropped A20 and A10 support from LE
<ndufresne> jernej: all I wanted is to test tiling ;-D
<ndufresne> I notice that overlay plane is can only scale opaque formats
<ndufresne> so if I pass BGRA, it will not scale, while if I passe BGRx it wil
<ndufresne> but haven't used/tested that base plane, as it used by fbcon
<jernej> I'm not really familiar with A20 display pipeline
<ndufresne> it's suni4
<ndufresne> no idea if there is variant withing that branch
<ndufresne> anyway, all I wanted is something to finish tiling to display integration, but if the decoder can't spit valid images mainline ...
<jernej> sun4i-drm driver support all socs from A10 to H6 and eventually newer
<jernej> currently 3 generations of display engine
<jernej> but first one, used on A20, is a bit weird
<jernej> composer part is split in two parts, called frontend and backend
<jernej> and only one can YUV formats
<jernej> IIRC
<jernej> ask mripard, he implemented support for this
<jernej> ndufresne: test LE image from link above, you can check if HW decoding works
<jernej> but this is older kernel...
<ndufresne> a bit unreliable, but at least decode in HW clean images in the image you privided
<jernej> unreliable in what way?
<ndufresne> this really looks downstream patches never made it, or some regression
<ndufresne> the UI most likely, it hangs, I had to stop/ play fiew time before it manages to start playing bbb 1080
<jernej> I didn't have any special patches for Cedrus on A20
<jernej> try 720p
<jernej> 1080p is probably on upper limit
<ndufresne> I see, but kind of work, wanted something hard enough so I'm sure it's not falling back to sw without telling ;-D
<jernej> press "o"
<jernej> you can see based on codec string if it is SW or HW
<ndufresne> ok, format drm_prime, I guess this is right
<ndufresne> and now i see that the CPU get maxed out for few s (about 10s) when I press play
Danct12 has quit [Remote host closed the connection]
<ndufresne> it's HW decode, but the two cpus are always around 50%, which is pretty high
Danct12 has joined #cedrus
<ndufresne> anyway
<jernej> yeah, kodi isn't the most gentle app
<jernej> lots of background tasks
<jernej> it may be possible to optimize a bit...
<ndufresne> jernej: so I pretty reproduce what gediz0x539 reported on A13, but I'm on A20
<ndufresne> this is using same gst build as I run on H3, and works all right on H3
<jernej> so, I can't really help here
<jernej> I don't have A13 nor A10
<ndufresne> I'm on A20, but I guess it's all the same line ..
<jernej> yeah, they all have similar display pipeline and VPU variant
<jernej> btw, A13 is single core cortex A8
<jernej> while A20 is dual core cortex A7
<ndufresne> oh well, I don't care enough to spend much time on it, I really just wanted to test something
<ndufresne> I got plenty of better SoC for my own entertainment ;-D
<ndufresne> but yeah, this is some 5.9 kernel
<ndufresne> looks like more of less cedrus from 5.10
<ndufresne> anyway, totally clueless on what may have broken it
<jernej> Exactly, cedrus should not be the issue
<jernej> maybe format configuration issue?
<jernej> or did it work before?
<jernej> since you already have this format supported in gstreamer