09:34 <mjourdan> xdarklight: Not really, I don't think I've ever used genpd. I used to have a meson-pwrc-vdec but since the blocks for H264 and HEVC require different settings in those registers I just ended up doing everything in the driver. It's only 2 registers so I didn't want to bother with a whole power driver.

09:35 <mjourdan> But if there's another way to do it so we can use the vdec driver on meson8 then sure

20:07 <xdarklight> mjourdan: I educated myself how to implement a power-controller. however, I'm not sure about the consumer side yet.

20:07 <xdarklight> mjourdan: do you have some Meson8b patches floating around somewhere? if you want I can do some experiments and post an RFC patch

20:08 <xdarklight> (your video decoder driver is probably "on hold" until the firmware license is finally cleared - fingers crossed that this will happen soon. let's also hope that the license covers the 32-bit SoC firmware files)

20:08 <Xogium> yay, kernel oops with nfsd on odroid c2 with 4.17 was fixed with 4.19 apparently

20:19 <mjourdan> xdarklight: I don't, sorry. Biggest issues I can think of are: meson8b doesn't support the "endianess" canvas field, so we're stuck with the tiled picture format (can't have NV12 out of the box). The HEVC code is also a bit different iirc. Otherwise porting the current driver to m8b should be "relatively" painless. I got H264 "working" back in the day on C1 so it really shouldn't be too hard.

20:21 <mjourdan> xdarklight: yes indeed the upstream status is kinda stalled right now, I'm currently focusing my efforts on getting userspace (kodi, chromium, mpv) to work as best as it can with the decoder; since most v4l2-related userspace code is still somewhat bleeding edge.

20:29 <narmstrong> mjourdan: not a big deal if on the other side we don’t need the endianned bit either

20:29 <narmstrong> We will only need to add a modifier on top of NV12

21:26 <mjourdan> narmstrong: It's a medium deal for egl dmabuf import since mali won't understand the format. But for drm yeah not much of a problem.

22:31 <xdarklight> mjourdan: OK, I guess a better approach (if I want to split the AO_RTI_GEN_PWR_... bits into a separate driver) would be testing on the GX SoCs. then I'm sure that I don't break your work

22:32 <xdarklight> mjourdan: what's the board you typically test with? I can test myself on a Le Potato, Khadas VIM1, Khadas VIM2 and some random other ones (I don't have any working GXBB board though)