02:48tertl3: LaserEyess: how is nouveau on it?
02:49tertl3: i have a 1060 but no no hardware decoding
02:49tertl3: i think we have same card with different name
12:57pmoreau: "FAILED 2 of 59 tests" (with 12 additional tests known to fail which were not run, so should be 14 out of 71) for basic.
13:34LaserEyess: tertl3: I've never tried nouveau on the 960, I was misinformed at the time I had it and I thought the proprietary drivers were my only option for a working computer
13:37 < mia\> hwdec wasn’t working for mine iirc but that was years ago
13:38 < mia\> but for a bunch of desktop stuff nouveau was actually more responsive than the nvidia driver because it didn’t introduce weird latency spikes and stall the X server every time something tried to init a gl context
13:39 < mia\> opening a web browser or mpv is very noticeably faster on nouveau
13:39LaserEyess: also the 1060 is absolutely not the same card as the 960, not even close
13:40 < mia\> especially if you try to use “weird” glx visuals, like transparent windows
13:42 < mia\> nouveau is also more stable when you try to use more than 3 displays. for some reason nvidia’s driver doesn’t like that, and you get all kinds of weird behaviors (vblank interval suddenly offset for one or more displays even if they all use the exact same modeline, random freezes)
14:08pmoreau: karolherbst and imirkin: Finally starting to get some Nouveau patches merged in: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10711.
15:37karolherbst: pmoreau: not 100% on some patches, so just a quick review for now
15:37pmoreau: Sure, no worries :-)
15:37pmoreau: Any feedback is good!
16:34imirkin: pmoreau: will try to review today
16:34pmoreau: Thank you!
16:34imirkin: LaserEyess: mia\: no vdec on maxwell and later.
17:53imirkin: pmoreau: hey, btw do you have a nva3+?
17:53imirkin: (plugged in and available to do stuff with)
17:54imirkin: if so, could you run https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10569 against dEQP-GLES31 on it for me?
17:54imirkin: in theory it should just be some texgather fails + some random compute/graphics interop issues
18:23pmoreau: imirkin: NVAC should count as NVA3+, right? If so, it’s definitely plugged into my laptop. I’ll try to run it most likely tomorrow.
18:35imirkin: pmoreau: don't think so
18:35imirkin: pmoreau: it has nva0 features, but not nva3
18:36imirkin: (specifically DX10.1 stuff like tex gather, texture query lod, probably some others)
19:29pmoreau: Mmh, then I’m not sure I have an NVA3+. I need to double check what my highest Tesla card is.
20:05imirkin: pmoreau: some macbooks had nvaf in them
20:05imirkin: which is nva3+
20:10imirkin: anyways, i'm moderately sure i have a nva8 somewhere
20:10imirkin: i'll swap it in for the g84
20:11karolherbst: pmoreau: I have an idea
20:12karolherbst: let's remove PIPE_COMPUTE_CAP_MAX_MEM_ALLOC_SIZE and let clover calculate something based on PIPE_COMPUTE_CAP_MAX_MEM_ALLOC_SIZE
20:12karolherbst: ehh...
20:12karolherbst: PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE
20:13karolherbst: those are purely clover flags anyway
20:13karolherbst: and I'd rather have every driver behace the same
20:14karolherbst: CL has some specific reqs on the max alloc size and I think it's legal to advertise 1/4 of max mem size
20:14karolherbst: or something
20:16pmoreau: I think we have that 1/4th computation somewhere already.
20:25karolherbst: mhhh