Daanct12 has quit [Remote host closed the connection]
pevik_ has joined #linux-msm
Daanct12 has joined #linux-msm
flto has quit [Ping timeout: 480 seconds]
Daanct12 has quit [Remote host closed the connection]
pespin has joined #linux-msm
<narmstrong>
Marijn[m]: ok I think I found the issue: the DSI link must be /3 because we transmit 8pp instead of 24bpp because of DSC, but we must still feed the MDP with the normal rate
<Marijn[m]>
narmstrong: I thought the original code already took care of that, or only for cmdmode?
<narmstrong>
Marijn[m]: seems Video mode has never been tested upstream because all this code is missing.. for cmd mode no need because you simply set mode->clock with hdisplay/3 and you use TE to do the vsync
<narmstrong>
so now I need to find which clock to keep 1:1 and which one 1:3
<Marijn[m]>
narmstrong: And for sanity check: if we don't use TE, should I still get proper frames albeit with tearing, or corrupted output?
<narmstrong>
and the DSI timings must be calculated on 1:3, this explains why the panel qcom,mdss-dsi-panel-phy-timings were totally off
<narmstrong>
Marijn[m]: on video or command mode ?
<Marijn[m]>
Asking because at some point I wasn't sure if my INTF TE support was proper, but I've tested it with multiple non-DSC devices now
<Marijn[m]>
*successfully
<Marijn[m]>
narmstrong: cmdmode
<narmstrong>
Marijn[m]: I thought TE was mandatory for cmd mode ?
<narmstrong>
(sorry I'm noob on command mode)
<Marijn[m]>
Oh about that: I have two devices with identical panel but different dsi-panel-phy-timings... sm6125 is all nice, sm8150 is half-fps until I bump up the porches
<Marijn[m]>
narmstrong: indeterminate. We've had TE break on non-dsc panels and either the mdp takes over (sync_cfg_height) or you just get nasty timeouts, tearing and slow panels, but it still "works"
<Marijn[m]>
I.e. setting sync_cfg_height to the panel height works without receiving TE from the panel
<Marijn[m]>
But doing it half rate is just slower (and with tearing iirc, if timed badly)
<narmstrong>
Marijn[m]: ok so if the frame size is correct, and if TE is out of sync I supose yes the output would be ok
<Marijn[m]>
We saw this with TE on PP because those registers are ignored, as we had to use them on INTF instead
<Marijn[m]>
Output was still working though
<Marijn[m]>
Maybe we should double-check if our dsi and mdp rates are correct for cmdmode too? I haven't really dabbled in that yet...
<narmstrong>
yep prob
<narmstrong>
on video mode, it explains why I have the top half of the display duplicated on the bottom half
<Marijn[m]>
Right yeah...
<narmstrong>
the display expects a full frame, but i receives only half the data
<Marijn[m]>
Don't think you caught my message last time: have you made any generic changes on top of my patch series that could affect cmdmode?
<narmstrong>
so I need to make the MDP run twice as fast..
<Marijn[m]>
Because I still get pure garbage as per the gitlab issue
<narmstrong>
Marijn[m]: no so far I don't have significant changes for cmdnode
<Marijn[m]>
(feel free to check that out and report on the magic status register!)
<narmstrong>
after the holidays I'll swicth to cmdmode
<Marijn[m]>
I hope to have the initial intf te up by then... Been busy clearing out my patch backlog lately
<Marijn[m]>
narmstrong: Pushed anywhere so that I can check them out? It could be anything small at this point
<narmstrong>
Marijn[m]: not yet
<narmstrong>
I must check I don't have any secrets before publishing the branch
<Marijn[m]>
No worries, I can only look at it tonight (if you can push it before the holiday at all)
<Marijn[m]>
Inb4 one of those secrets fixes it
<Marijn[m]>
I don't mind if you push them though 😏
<konradybcio>
qualcomm legal would :P
flto has joined #linux-msm
flto has quit [Ping timeout: 480 seconds]
krzk has quit [Remote host closed the connection]
krzk has joined #linux-msm
pespin has quit [Remote host closed the connection]
flto has joined #linux-msm
sibis has quit []
sibis has joined #linux-msm
sibis has quit []
sibis has joined #linux-msm
pevik_ has quit [Ping timeout: 480 seconds]
* Mis012[m]
should be making more use out of his knowledge as someone who didn't sign an NDA with qcom to obtain it
<aka_[m]>
Mis012: you shoudl
<Mis012[m]>
I ordered the easy mode sdcard extender, hopefully will make it harder to procrastinate NIDnT
<Mis012[m]>
aligning the pcb in the combo slot is much more of a PITA than I thought