09:06pmoreau: karolherbst: I believe that error (the vloadn one) was fixed in SPIRV-Tools.
10:25karolherbst: pmoreau: mhh, which version? I have 2019.4 here
10:28pmoreau: You need https://github.com/KhronosGroup/SPIRV-Tools/commit/13f61bf859446ca5b0e9cf6260ba24ca7e30ed09, which seems to be in 2019.4
10:28karolherbst: yeah, then it doesn't work
10:28pmoreau: But the error message was updated too and you are still seeing the old one.
10:28karolherbst: which is od
10:28karolherbst: maybe I have a stale and old spirv tools somewhere
10:29pmoreau: You should be seeing “expected operand P storage class to be UniformConstant, Generic, CrossWorkgroup, Workgroup or Function“ now.
10:29pmoreau: Yeah, maybe
10:29karolherbst: mhhhh... I have one in my local installation, but I think it was 2019.4 as well...
10:29karolherbst: oh well
10:30pmoreau: Or was it the development version of 2019.4, and not the released version? So you might have built it before the patch went in.
10:30karolherbst: let me check
10:31karolherbst: now, should be part of .4
18:55Lyude: a-ha, knew there had to be something wrong, you -can- read the CRC values before doing a notifier swap
18:56Lyude: although the experiment I used to figure this out I could have sworn I had already tried a few times, weird.
18:57imirkin_: Lyude: phase of the moon changed, so should be much better now
18:57Lyude: can't spend much time on this today, but this discovery makes things much easier :)
19:02diogenes_: Hey guys, does anyone have any info on how nouveau runs on netbsd and if any work is still ongoing in that direction?
19:04imirkin_: apparently they're up to at least the code in 4.19, which should include support for many GPUs
19:06imirkin_: but even 3.15 included maybe even maxwell, but def all kepler gens
19:10diogenes_: sorry, got disconnected.
19:13imirkin_: diogenes_: check logs
19:13diogenes_: imirkin_, last line: <diogenes_> which means the work stopped right there i guess.
19:13imirkin_: check logs for what actually happened :)
19:14imirkin_: see topic for logs link
19:17diogenes_: ohhh nice, got the logs, thanks guys.
19:17diogenes_: is DRI_PRIME a nouveau thing or X?
19:18imirkin_: it's an X thing and also wayland, i think
19:18imirkin_: but each driver also needs to support it
19:18imirkin_: a lot of things have to work together to make it happen
19:19diogenes_: imirkin_, i see, thanks.
19:34karolherbst: actually, DRI_PRIME is a mesa thing
19:35karolherbst: and nvidia implemented it with different env variables, so...
19:56Lyude: yay, CRCs output consistently now. almost certainly need to fix us missing a few crc events due to forgetting to flush entries that got written to a notifier ctx after we flipped it, but that's quite easy to fix. Interestingly, I believe (at least for rg and sf, no idea about outp yet because I still need to figure out what's wrong with it) that crcs might actually take two CRC entries per frame instead of
19:57Lyude: Since I'm seeing CRC values always alternate between two values each entry, even when getting the CRC of a static image
20:40imirkin_: Lyude: dynamic dithering?
20:40Lyude: imirkin_: I was just about to say maybe it could be that
20:40imirkin_: you can configure the dither mode
20:40Lyude: since I just noticed they mark the dithering bit in the documentation
20:41imirkin_: we default to dynamic_2x2, but you could do off
20:41Lyude: imirkin_: mhm, i'll give that a try next time I'm looking at this
20:42imirkin_: were you on a lvds panel?
20:42imirkin_: (or eDP or whatever)
20:42imirkin_: lots of 6bpc in there
20:42Lyude: imirkin_: nope, this is mst actually
20:45imirkin_: not sure that's mutually exclusive, but i presume you have a normal monitor attached too
20:45Lyude: imirkin_: "normal" == ?
20:45imirkin_: not a laptop panel
20:45Lyude: ah yes, it's not a laptop panel
20:46Lyude:shudders at the thought of an mst panel
20:46imirkin_: well, it's DP ... just need to get the wiring right :)
20:46imirkin_: requires steady hand. will void warranty.
20:55karolherbst: Lyude: I am wondering how dual display laptops will be implemented if not with MST :O
20:58karolherbst: Lyude: you know this Asus ZenBook Pro with two screens? I am actually wondering what they do internally
20:58Lyude: karolherbst: eDP
20:58karolherbst: on both displays?
20:59Lyude: karolherbst: yep, same eDP connection as well iirc? I don't actually have access to the spec but I'm pretty sure that's the idea
20:59Lyude: was something samsung asked to get into the edp spec iirc
21:00karolherbst: OEMs should stop screwing around and start creating proper laptops again...