00:22 imirkin: skeggsb: any strong objection to flipping mcp89 to mcp77_mmu instead of g84_mmu?
00:22 imirkin: i don't think compression ever worked with nouveau
00:22 imirkin: and unclear whether it's possible in the first place or not
04:20 AndrewR: imirkin, around?
04:20 imirkin: sorta
04:31 AndrewR: imirkin, should I 'formalize' this shader-based mjpeg decoding request, via feature bug for mesa? Or email to nouveau ... Don't want to create 'useless' bugs, but I had hard time re-finding this cuda based _open source_ de/encoder .... Google changes ....
04:31 imirkin: if you want a shader-based mjpeg decoder, you can file a feature request for it
04:31 imirkin: it would be part of vl, not nouveau
04:33 AndrewR: ok. (but I was thinking about compute support, you said some time ago opencl images for my nv50 card not really implemented, but eventually they may ...so, some additional GPU decoding can be done on card then? But it will be soewhat NV specific?
04:34 AndrewR: I mean, if current implementation in this library I found depend on cuda..it may work for other cards, too..if used from layer below. But hw is different, so problems may surface (low speed, etc)
04:45 imirkin: should still be a common impl
04:52 AndrewR: imirkin, https://bugs.freedesktop.org/show_bug.cgi?id=110699
05:29 AndrewR: imirkin, it seems CESNET's version uses CUDA for _everything_ .... but then CUDA as API just uses underlaying hw, and jpeg_gpu was still working for me on nouveau without any compute support :} Also, I noticed they allowed usage of clang as cuda compiler.. but into what exactly it will compile? (q. to myself, mostly)
05:31 AndrewR: https://stackoverflow.com/questions/12099684/compiling-cuda-with-clang -"Note that using Clang for compiling CUDA still requires that you have the proprietary CUDA runtime from the NVIDIA CUDA toolkit installed."
05:32 pabs3:wonders if coriander works with nouveau https://github.com/hughperkins/coriander
05:44 AndrewR: pabs3, probably not yet :} (I mean it may, if karol herbst will enable everything for opencl (1.2?) at least on nvc0 ...)
06:26 HdkR: Woo for TU117 support in 5.2? :D
10:52 RSpliet: HdkR: modesetting yes ;-)
12:29 HdkR: RSpliet: Good enough for an expensive display output :P
17:57 Lyude: karolherbst: whenever you get a chance (no rush since I assume you must be tired from your flight back :p) do you think you could get me a copy of the patches that you said would fix those secboot issues after runtime PM on pascal? One of the people I had test that patch series seems to be hitting timeout issues with their GPU, although I have no idea if there's any relation with the secboot after runtime PM
17:57 Lyude: issues
17:57 Lyude: but i assume we should probably at least try that patch series first before trying to debug it any further
17:58 karolherbst: Lyude: there should be on my secboot_fixes branch in the nouveau tree
17:58 Lyude: karolherbst:gotcha
18:31 karolherbst: Lyude: but with the runpm stuff finally working, battery life time is insane
18:31 karolherbst: after the flight I got still 40% left and it was like 7 hours
18:32 Lyude: hehe, I'd imagine
18:32 karolherbst: I mean I was doing the ACPI stuff before that
18:32 karolherbst: but it seems like it wasn't saving as much power
18:33 karolherbst: Lyude: ohh, and it seems like to work with system suspend/resume just fine as well
19:04 Lyude: karolherbst:yep :), I saw the same thing on my zbook