12:03 <Marijn[m]> vkoul: calebccff: can any of you retest the Pixel 3 panel with my DSC patch series? It's now become a requirement before merging :)

12:03 <vkoul> Marijn[m]: I no longer have that device :-( maybe calebccff can

12:05 <Marijn[m]> vkoul: that's unfortunate, then calebccff is my only hope

12:07 <Marijn[m]> vkoul: any other DSC device you might be able to test? Are you still working on DPU?

12:08 <Newbyte> Marijn: did you update the DSC patches for sdm845?

12:08 <vkoul> Marijn[m]: not right now, sorry!

12:08 <Newbyte> I have an sdm845 device with DSC but I could never get anything usable out of the panel (this was before the DSC patches were merged)

12:11 <Marijn[m]> Newbyte: this patch series is specifically to resolve dsc on sdm845 :) - do see my comments about updating the dsc configuration in the panel driver though

12:11 <Newbyte> Marijn: maybe it's time to re-test then. and, what comments are you referring to?

12:12 <Marijn[m]> vkoul: that's unfortunate, anyone else that is? I'm working on DPU on and off and would hate to do overlapping work

12:15 <Marijn[m]> Newbyte: what device and panel are you using, can you link me the driver?

12:15 <Newbyte> Marijn: it was just generated using this: https://github.com/jenneron/linux-mdss-dsi-panel-driver-generator/commit/47d4ebca241dbd400ce7c8d9ec8051b7477f7deb

12:16 <Newbyte> and the device is LG G7 aka lg-judyln: https://github.com/torvalds/linux/blob/master/arch/arm64/boot/dts/qcom/sdm845-lg-judyln.dts

12:16 <Marijn[m]> Newbyte: The comment I'm referring to is https://lore.kernel.org/linux-arm-msm/20221001214816.6l3yo4bgwmfcy3gr@SoMainline.org/, but you can also use https://github.com/SoMainline/linux/commit/86fa35ff2499f4f2600456caf80139e47b54f4b5#:~:text=dsc as a reference, just make sure to line up everything that says `dsc`

12:17 <Marijn[m]> Newbyte: Oof yeah that generator is wrong 😅

12:17 <Newbyte> great to hear

12:17 <Newbyte> do you have anything better to suggest?

12:21 <Marijn[m]> Newbyte: A PR from jenneron and then I add some review comments - or I take over the branch and apply some fixes to it? It's not much, just the hardcoded slice count and `bits_per_pixel` mismatch

12:22 <Newbyte> Marijn: jenneron is busy right now (and he's in Ukraine), so I think he'd prefer the latter. I can ask though

12:23 <Marijn[m]> Newbyte: Sure! Now that we've got DSC "working" upstream lowering the barrier with panel-driver-generator support is definitely preferable IMO

12:23 <Marijn[m]> "working" as in, sdm845. The newer SoCs are still having trouble but rest assured it's being investigated :)

12:24 <Newbyte> Marijn: if you fix those things with the panel generator please ping me and I'll be happy to test it on the G7

12:25 <Newbyte> if you want any clarification from Jenneron that you're allowed to use his work I can ask

12:25 <Marijn[m]> Newbyte: Do you already have a driver that I can look at? It's really only adjusting the `bits_per_pixel` to `8 << 4` and booting with my patches applied of course :)

12:26 <Newbyte> I'm not sure … it was like a year ago

12:26 <Marijn[m]> Newbyte: Might be nice but I hope jenneron is fine with it if I base on top and maintain (co-)authorship that way

12:27 <Newbyte> Marijn: he said "yes of course"

12:29 <Newbyte> (with regards to whether he's okay with that)

12:30 <Marijn[m]> Newbyte: Perfect, I'll make the changes tonight and let you know!

12:31 <Newbyte> Sounds good to me :)

14:07 <calebccff> Marijn[m]: I'll give it a go!

14:07 <calebccff> can you drop a link to the series? :)

15:36 <Marijn[m]> calebccff: it's in the comment for your panel driver as well

15:37 <Marijn[m]> calebccff: alas, the link from lore frontpage: https://lore.kernel.org/linux-arm-msm/20221009184824.457416-1-marijn.suijten@somainline.org/T/#t

15:38 <Marijn[m]> Again, follow the comment in your panel driver to make it compatible. I'll push a panel generator fix shortly

15:56 <Marijn[m]> calebccff: Newbyte I just pushed my version of the panel generator to https://github.com/MarijnS95/linux-mdss-dsi-panel-driver-generator/tree/dsc

15:57 <Marijn[m]> Disclaimer: hacked this together in the train based on a local copy of my LGD Akatsuki driver. No testing of the generator nor its output has been done!

16:07 <Newbyte> Marijn: Thanks, I'll try it tomorrow hopefully

16:19 <calebccff> Marijn[m]: what kernel base did you use? this doesn't apply cleanly on 6.0

16:21 <Marijn[m]> calebccff: -next, with Dmitry's dsc changes

16:22 <Marijn[m]> I think rebased on -next from ~4 days ago, though there haven't been changes in the past ~4wks

16:22 <calebccff> alrighty, i'll switch to that

18:19 <calebccff> Marijn[m]: rebased, applied your patches, re-generated the driver and fixed up the generator, no luck still: https://gitlab.com/sdm845-mainline/linux/-/commits/caleb/pixel3-bringup-next-2022-10-13

18:21 <Newbyte> calebccff: how did you have it working eralier then?

18:21 <Newbyte> s/eralier/earlier/

18:21 <calebccff> Newbyte: was using bootloader configuration

18:22 <calebccff> the driver I was using was pretty heavily modified, and was missing a bunch of init stuff. With the re-generated one I just get a black screen if I reset the panel, if I don't reset the panel I get a bunch of dark noise stuff

18:22 <Marijn[m]> "No luck still" so it's a bit weird that people on the mailing list are pretending it was properly working before :)

18:22 <Marijn[m]> calebccff: Can you comment out the `picture_parameter_set` call for the time being, just to hope that the panel is left in the same state?

18:23 <Marijn[m]> calebccff: changes on your branch look good though 😅

18:24 <Marijn[m]> I used to compare the logs from that `"DSC params:"` hex print between downstream and mainline, if that's useful

18:25 <Marijn[m]> calebccff: In case of a black screen, I almost don't trust the backlight code in that driver - Sam commented on it before too. Are you sure this panel only uses 8-bit? Most ones I have here use at least 10 but sometimes even 12 bits.

20:16 <calebccff> Marijn[m]: well the patches I sent had it working... you just had to skip the reset sequence

20:16 <calebccff> will try your suggestion

20:16 <calebccff> I messed with the backlight too, including switching back to the backlight code I had before which definitely worked

20:17 <calebccff> and disabling backlight altogether

20:17 <calebccff> my damn usb-cereal UART adapter has given up on me, so I can't get anything out of the device, will have to toy with usb serial gadget

20:27 <Marijn[m]> calebccff: So my patches, with your newer driver, with the reset commented out (which is typically tricky to get right considering multiple command sequences, gpios to poke, and multiple codepaths that result in other commands being sent downstream), make DSC work? With or without sending PPS over DCS to the panel?

20:28 <Marijn[m]> calebccff: Curious if you can get backlight working - over the full range of the panel - without the reset too. Pretty easy to bump the maximum value to like 1024 or 4096, and see if the emitted brightness "restarts" every 1/4th of the `brightness` sysfs attribute

20:29 <Marijn[m]> These panels are all a bit silly in requiring a big-endian value, and Sam commenting that it "should be handled internally". We've needed it on every panel thus far 😬

20:29 <calebccff> Marijn[m]: I couldn't get it working at all with the new driver

20:30 <Marijn[m]> calebccff: Feel free to switch to your original driver, insert the `* << 4` and keep the PPS out of it, see if it works. From there on we can start enabling PPS and go from there?

20:34 <calebccff> sounds like a plan, will try and take a look tomorrow