03:15imirkin: skeggsb: in case you're around ... how do you trace blob display stuff?
03:15imirkin: i mean the evo channel specifically
03:16imirkin: do you mmt Xorg? or something else clever (esp with modern blob, not sure where it all happens anymore)
03:16skeggsb: you can't really anymore, not easily at least.. the kernel driver does it now, and the pushbuf is in system memory
03:17imirkin: hrmph ok
03:18imirkin: do you happen to know the cut-off?
03:18skeggsb: not sure what version they switched to kernel modesetting
03:19skeggsb: pascal will be ok with mmiotrace, pushbufs are forced into vidmem there for a hw limitation
03:19imirkin: i kinda need to do this on kepler
04:14imirkin: HdkR: do you know if there's a way to force blob to do yuv420 even if hdmi 2.0 is available?
04:15HdkR: Is it not exposed in nvidia-settings?
04:15imirkin: no clue
04:15imirkin: preparing for the future :)
04:16imirkin: i haven't run blob in *ages*
04:16imirkin: hopefully i can get it to run without destroying my system
04:16imirkin: but you seem to know odd things like this, figured i'd ask
04:17HdkR: Yea, looks like it is in nvidia-settings
04:17HdkR: https://forums.developer.nvidia.com/t/rtx-2080-ti-unable-to-achieve-hdmi-4k60-with-444-chroma-subsampling-under-centos/108962 According to the response here
04:17HdkR: Been about two years since I've run the blob myself
04:17imirkin: hah ok
04:18imirkin: interesting --
04:18imirkin: looks like there's an xorg.conf setting for it
04:19HdkR: Yea, nvidia-settings tends to just drop things in xorg
04:19imirkin: well, this makes it sound like there's no option in nvidia-settings for yuv420
04:19imirkin: only rgb/yuv444/yuv422
04:19HdkR: ah, missed that
04:20imirkin: or was the case 4 years ago
04:23imirkin: skeggsb: ok. so i'm going to try this with pascal + mmiotrace. any clever suggestions for how to track down where in the mmiotrace the vram pushbuf writes are?
04:50skeggsb: imirkin: grep for the PUT reg for the channel you're looking at
04:51imirkin: ah yeah. right.
18:37imirkin: karolherbst: when you get a chance: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8901
18:37imirkin: passes the deqp scissored clear stuff
18:38imirkin: (which helped identify some earlier issues, with out-of-bounds scissors)
18:38karolherbst: mind mentioning the deqp regex so I can just run it without having to figure it out myself?
18:38imirkin: well, there's probably something more precise, but i did *clear* :)
18:38imirkin: (in gles3)
18:39karolherbst: yeah.. but I can give it a try on turing at least
18:39imirkin: well, this is only tested on pascal and tesla. but this isn't like ... frequently changing stuff