06:47 tomeu:googled for rockchip's tensor format of choice, and found one more REing effort in github
06:47 tomeu: https://github.com/phhusson/rknpu-reverse-engineering
07:37 f_: hey that one is from phh!
07:42 tomeu: ah, you know him?
07:46 f_: a bit, yes
07:46 f_: He's in #linux-amlogic
09:15 Danct12: let's invite him here
09:18 tomeu: Jas' effort seems to have been based on top of him's
09:53 f_[xmpp]: Danct12 Sure
12:11 tomeu: daniels: didn't go that well :( https://gitlab.freedesktop.org/mesa/mesa/-/jobs/55123716
15:53 tomeu: tried loading etnaviv as a module later in the boot, but that ended up working even worst: https://gitlab.freedesktop.org/mesa/mesa/-/jobs/55135661
15:53 tomeu: as the machine hung everytime I loaded etnaviv
15:54 tomeu: though, looks like in the same pipeline, the gitlab-ci job was retried automatically, and it worked at the first go: https://gitlab.freedesktop.org/mesa/mesa/-/jobs/55137098
15:54 tomeu: (on cbg-0)
16:01 daniels: I wonder if we need to upgrade the bootloader, because they all have the same power supply now :\
16:01 daniels: and it seems unusual that modprobing etnaviv would cause so much load that it would immediately take it out before anything could even be printed to uart
16:01 tomeu: yeah, looks like it
16:01 daniels: so I'm thinking more regulator/clock issues
16:02 tomeu: yeah, configured by one of the BL* blobs
18:43 f_: tomeu: have you got UART logs? Curious.