01:15anholt: karolherbst: making progress on standing up CI for jetson nano.
01:16anholt: I've got the power control working in a test environment, now trying to migrate all the setup to gitlab-ci.
01:16anholt: I'd be interested in suggestions for any other boards besides jetson nano that would be interesting to test with and can power off of <=15w of barrel jack or usb.
02:47juri_: anholt: I'd be interested in a guide on adding a board/system to CI.
02:50juri_: I can probably add at least 5 cards.
06:43anholt: juri_: for nouveau, the easiest strategy I think (and we've done a few strategies between different boards so far, so I think I've explored the space well) is a managed POE switch, POE splitters for the boards, serial per board, and an x86 runner device talking to the switch and the board's serial, with probably like 3gb of ssd storage per board plus some caching.
06:44anholt: there are some docs on docs.mesa3d.org, but the poe strategy isn't really documented. I'll probably write some in the process here.
13:57imirkin: anholt: there's also jetson tk1 which covers kepler (iirc nano is maxwell?)
13:57imirkin: i have one, but i never was able to get it to work properly with my own kernels
13:57imirkin: network activity seems to cause it to fall over, which isn't a great combo with nfsroot
19:35imirkin: anholt: also some generaction of imac mini (or whatever their NAS-like device was called) have nvac (aka MCP79)
20:32HerrSpliet: imirkin: I may or may not have an NVAC (Asus, Intel Atom 330, NVIDIA ION) that I won't be using anymore
20:34RSpliet: There's the slight challenge that currently that machine is in the Netherlands, whereas I am not
20:34RSpliet: And last time I wanted to bring it back the "host" couldn't find it
20:34RSpliet: So... may or may not
20:37RSpliet: anholt: ^
22:44mynacol: imirkin: I think I did it!
22:46mynacol: I fired up a Linux Mint and did a mmiotrace of nvidia-340. The instructions of the web page resulted in fuc409c/d and fuc41ac/d
22:47mynacol: I moved them at the right place (I guess), but I have no proof that they actually work. Should there be a log entry in dmesg?
22:48mynacol: Once I included nouveau in the initrd and an error that the firmware files could not be found was printed, so I guess my NvGrUseFW=1 works...
22:58mynacol: So far no freezes with mpv, so I am quite pleased :)
22:58mynacol: Thanks for the help
23:07mynacol: My first attempt was not to mmiotrace, but to modify extract_firmware.py to directly get the firmware. Would be much easier for future people. No real progress from me though.