04:14imirkin: kodomo: the reason you get 2D_NVC0 fail is because the whole accel unit fails to come up on that board
04:15imirkin: kodomo: however might i ask... what's your goal? there are no outputs on there, it's just an accelerator -- you're better off just using the intel gpu.
04:15imirkin: gnarface: lspci don't lie. if it says GP108, it's a GP108.
04:16imirkin: kodomo: there have been recent efforts at getting the runpm code to play nice with the GP108, you're best off getting a fresh kernel and trying again
05:54mupuf: I can confirm that the gitlab vbios repo is up to date, so it can be used to add more
05:54mupuf: however, the name is really attrocious
05:57mupuf: the shaderdb one is also up to date
08:34tjaalton: imirkin: to fix asserts in mesa 19.0.x I'd probably want "23dfff0669ef351 nouveau: flip DEBUG -> !NDEBUG"?
08:35tjaalton: "fix" as in not hit them
10:32tagr: mooch2: from a brief search I can't find anything useful on those devices, but I think the nouveau community has done a great job of documenting these
10:33tagr: mooch2: for older chips you may want to look at the nv driver, which has some support for pre-nv04 I think
10:34tagr: mooch2: I can ask around if anybody knows of some internal information about these, but frankly, these are so old it's unlikely I'll find anything, so don't get your hopes up
10:35tagr: mooch2: it might help if you have specific topics that you want information on
11:42mooch2: tagr, i mostly need more info on pfifo
13:08imirkin: tjaalton: that change was to actually make the debug stuff show up more often, so you probably don't want it for a release build anyways
13:08imirkin: tjaalton: unfortunately the meaning of "DEBUG" has been co-opted by nir validate to mean "super-expensive checks you want when hacking on core stuff"
13:09imirkin: which means all previous uses of DEBUG have to change.
13:10imirkin: tagr: in case it wasn't made clear, mooch2's goal is to get the windows driver to work with the emulated hw.
13:12imirkin: tjaalton: for a release build you want to make sure you're not setting DEBUG, and *are* setting NDEBUG.
13:16tjaalton: imirkin: which was already happening, yet it hit asserts
13:17imirkin: clearly not then.
13:17tjaalton: clearly was, according to the build logs
13:18tjaalton: doesn't matter
13:18imirkin: for the specific version that the user had?
13:18imirkin: i thought you had enabled it for later builds or something
13:18imirkin: so the actual issue there wasn't that the assert() was false, but that the code inside the assert was crashing
13:19imirkin: however if asserts are disabled, i'm pretty sure that code shouldn't run in the first place
13:19tjaalton: what was the fdo bug again?
13:19imirkin: my memory is good, but ... not that good :)
13:19tjaalton: guess I found it
13:20imirkin: and its duplicate
13:20imirkin: dunno which is which
13:21imirkin: yeah, the dup was using a proper package - https://bugs.freedesktop.org/show_bug.cgi?id=111218
13:21imirkin: libgl1-mesa-dri: 19.0.2-1ubuntu1.1~18.04.2
13:22tjaalton: right, that's built on 18.04 which has meson 0.45.1, which fixed b_ndebug (was broken in 0.45.0)
13:23tjaalton: and inherited -Db_ndebug=true from debian
13:23imirkin: is it possible to view what (gcc) command these things are ending up getting built with?
13:24tjaalton: sure, for the latest version which is 19.0.8
13:24imirkin: i'm sure it hasn't changed too much
13:24tjaalton: ah, found the real one https://launchpadlibrarian.net/433791967/buildlog_ubuntu-bionic-amd64.mesa_19.0.2-1ubuntu1.1~18.04.2_BUILDING.txt.gz
13:26imirkin: can you point me to where NDEBUG is being set?
13:26tjaalton: why isn't it getting it then?
13:26imirkin: not my department.
13:26imirkin: i like automake
13:26imirkin: and autoconf
13:26tjaalton: so a meson bug
13:27tjaalton: wasn't autoconf dropped already
13:27imirkin: it was.
13:27imirkin: maybe not for 19.0 though
13:27imirkin: definitely for the upcoming 19.2
13:29tjaalton: seems to work right in 19.1.4
13:30imirkin: dcbaker (available in #dri-devel usually) is the meson expert.
13:30imirkin: i'm the anti-meson naysayer
13:34tjaalton: looks like something missing from meson 0.45.1
13:34tjaalton: 19.0.x built on 19.04 seems fine
13:34tjaalton: ok, I'll backport whatever
13:36imirkin: fwiw it's not just nouveau, there's no mention of NDEBUG in that whole build log
14:47tjaalton: i know
23:17Lyude: skeggsb: at some point we should implement DSC in nouveau
23:18HdkR: Lyude: Is there hardware that supports DSC yet?
23:18Lyude: HdkR: I mean, drivers/gpu/drm/drm_dp_dsc.c, amd and intel both have support for it in their driver and are adding MST DSC soon
23:18Lyude:is helping review those patches :)
23:18Lyude: HdkR: so yeah, there definitely should be
23:18HdkR: er, displays I mean
23:19Lyude: no idea, but maybe I can find one anyway
23:19HdkR: I haven't found any when I was looking, maybe it is more common in laptop displays over eDP?
23:19Lyude: HdkR: i dunno
23:19Lyude: i'll have to ask intel how they test DSC
23:20airlied: HdkR: didn't you get an 8k monitoir?
23:20airlied: I thought dsc was needed for 9k
23:20HdkR: It doesn't support DSC as far as I'm aware
23:20HdkR: It uses dual DP cables
23:20skeggsb: i don't think even turing supports it yet tbh
23:21skeggsb: at least, it doesn't report support in the head caps register
23:21Lyude: skeggsb: I definitely believe turing supports it, mdnavare mentioned it during xdc last year didn't they?
23:21skeggsb: #define NV_PDISP_FE_SW_HEAD_RG_CAPA_SUPPORT_DSC 16:16 /* RWIUF */
23:21skeggsb: 64.608177 write32 #3 +0x00640048 <- 0x00000a00
23:22airlied: HdkR: is it 8K@30?
23:22airlied: for @60 it needs dsc
23:22skeggsb: hmm, nvidia's website disagrees, it's possible the cap moved or something on turing, that's from volta's header
23:23HdkR: Although it is a 10bit panel that only exposes 8bit colour depth
23:25HdkR: airlied: It does some sort of wacky CRTC bonding like the old dell monitors in order to hit 8k60 :D
23:25Lyude: HdkR: tiling
23:25HdkR: That works
23:26HdkR: Bit smarter than the old ones though. If you pull one of the cables it drops to 8k30 instead of half the screen vanishing
23:26Lyude: oh neat
23:26Lyude: usually when i pull the display cables on my dell monitors they just crash :)
23:26Lyude: (not all of them, but too many of them...)
23:27Lyude: granted, most of my dell mst displays are very very early revisions
23:27HdkR: ah. I haven't done any MST since I don't trust it
23:27Lyude: HdkR: if mst ever breaks you can poke me ;)
23:27HdkR: I have two 4k monitors upstairs that support MST. Could give it a try sometime
23:28HdkR: I...don't even want to try the 8k monitor in Linux
23:29HdkR: It's connected to my gaming PC instead :P