00:29airlied: Lyude: enable CONFIG_RUST_BUILD_ASSERT_ALLOW
00:38fdobridge_: <gfxstrand> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27024
00:39Lyude: airlied: thanks :)
00:42Lyude: mh - yeah it does look like I need to implement something for platform devices
00:44airlied: I think asahi has some platrom bits
00:48Lyude: gotcha!
01:04fdobridge_: <esdrastarsis> GPL? Amazing! 🎉
01:18fdobridge_: <gfxstrand> And ESO
01:24fdobridge_: <phomes_> I did some testing with the usual games and demos. Everything is working well, and baldurs gate 3 is loading further than before (but still hangs)
01:28fdobridge_: <bylaws> Does baldurs gate 3 require gpl?
01:37fdobridge_: <gfxstrand> Do we have a D3D trace of BG3?
01:43fdobridge_: <phomes_> I wanted to provide some traces but I really do not know what I am doing for that. Would need to take a closer look
01:44fdobridge_: <phomes_> I have tested with both vulkan and directx11. Both hang at the same spot
01:45fdobridge_: <phomes_> fifo:000000:000c:[bg3.exe[36398]] rc scheduled
01:45fdobridge_: <phomes_> fifo:000000: rc scheduled
01:45fdobridge_: <phomes_> fifo:000000:000c:000c:[bg3.exe[36398]] errored - disabling channel
01:45fdobridge_: <phomes_> bg3.exe[36398]: channel 12 killed!
01:45fdobridge_: <phomes_> bg3.exe[36398]: error fencing pushbuf: -19
01:46fdobridge_: <phomes_> `fifo:000000:000c:[bg3.exe[36398]] rc scheduled
01:46fdobridge_: <phomes_> fifo:000000: rc scheduled
01:46fdobridge_: <phomes_> fifo:000000:000c:000c:[bg3.exe[36398]] errored - disabling channel
01:46fdobridge_: <phomes_> bg3.exe[36398]: channel 12 killed!
01:46fdobridge_: <phomes_> bg3.exe[36398]: error fencing pushbuf: -19` (edited)
01:46fdobridge_: <phomes_> and
01:46fdobridge_: <phomes_> `fifo:000000:000c:[bg3_dx11.exe[52555]] rc scheduled
01:46fdobridge_: <phomes_> fifo:000000: rc scheduled
01:46fdobridge_: <phomes_> fifo:000000:000c:000c:[bg3_dx11.exe[52555]] errored - disabling channel
01:46fdobridge_: <phomes_> bg3_dx11.exe[52555]: channel 12 killed!
01:46fdobridge_: <phomes_> bg3_dx11.exe[52555]: error fencing pushbuf: -19`
01:47fdobridge_: <phomes_> I can provide .foz files from ~/.steam/debian-installation/steamapps/shadercache/1086940/fozpipelinesv6 if that is useful?
02:32fdobridge_: <nanokatze> nothing requires gpl
02:46fdobridge_: <gfxstrand> "error fencing pushbuf" is funky...
04:11i509vcb: makinbacon12: yeah my hardware is ga10b
04:54fdobridge_: <makinbacon21> Ofc
04:54fdobridge_: <makinbacon21> T234 is ga10b
04:54fdobridge_: <makinbacon21> I've got agx orin
04:58i509vcb: ah okay so I have the smaller brother the orin nano
05:08fdobridge_: <makinbacon21> nano is nice
05:08fdobridge_: <makinbacon21> I like the debug port on the agx
05:08fdobridge_: <makinbacon21> and the PCIe slot
05:08i509vcb: I do have a stupid idea how to use the nano
05:09i509vcb: but that's going to take a few months to get arranged, so why not screw around with nouveau
05:10i509vcb: there's also someone who is making a DIY XR headset using the nano
05:11fdobridge_: <makinbacon21> I have a project I'd like to use mine for, in the meantime I've been doing stuff like getting lineageos to boot and yea now looking at nouveau
05:11i509vcb: I heard a comment online that someone said it plays games dolphin really well :)
05:11fdobridge_: <makinbacon21> oh I'm sure
05:11i509vcb: err under dolphin
05:12fdobridge_: <makinbacon21> It runs Rosetta too which is cool
05:12i509vcb: rosetta? wrong OS I think?
05:12fdobridge_: <makinbacon21> Nope
05:12i509vcb: FEX probably
05:12fdobridge_: <makinbacon21> Nope
05:12fdobridge_: <makinbacon21> Well that too
05:12fdobridge_: <makinbacon21> But
05:12fdobridge_: <makinbacon21> Apple added Rosetta 2 for Linux vms on Apple silicon devices
05:13i509vcb: the CPU is sadly not as great as my M2 air, but the GPU is half the reason I bought it
05:13fdobridge_: <makinbacon21> And it turns out with a couple simple-ish hacks you can get rid of some checks and run it on any a78-ae or higher cpu
05:13fdobridge_: <makinbacon21> So you can use it
05:14fdobridge_: <makinbacon21> I spent very little time with it but I did fire it up
05:14i509vcb: I'd be a bit concerned about the perf issues without having the M1's TSO extensions
05:16i509vcb: although super modern arm64 hardware on linux always brings up fun problems
05:16i509vcb: gdb and kicad had a problem with pointer authentication
05:17i509vcb: I'll probably try to start doing random stuff next week with seeing how painful it would be to get T234 under nouveau. Not really at home this weekend so I'll probably be implementing async flips on asahi instead
05:42fdobridge_: <makinbacon21> I've been fiddling with kanging gp10b
05:42fdobridge_: <makinbacon21> and adjusting
05:42fdobridge_: <makinbacon21> but once i started hitting stuff that ga100 uses GSP for it got weird
05:45fdobridge_: <airlied> probably easier to fork from the previous tegra
05:45fdobridge_: <airlied> not the gsp ga stuff
10:28montjoie: hello just a quick update on my problem I got some weeks go on a GP108 [GeForce GT 1030], with firmware I successfully loaded nouveau and got a working X https://paste.debian.net/1303874/
10:28fdobridge_: <redsheep> I have successfully grabbed a bg3 trace on windows and confirmed it replays there:
10:29fdobridge_: <redsheep>
10:29fdobridge_: <redsheep> https://mega.nz/file/h9RRWSiL#dS6NbR6kLngwKEmSMordu3__6UOXS-0K0G2ypa4R60M
10:29montjoie: unfortunatly performance for gaming is not good (but from a small distance, eu4 got a small lagging)
10:30fdobridge_: <redsheep> Unfortunately my attempts to replay that on nvk to try to replicate the error hasn't been so fruitful, it doesn't show anything and doesn't kick out the error
10:37fdobridge_: <redsheep> However, I can confirm that I can replicate the error on ada with 27024 applied. Here's dmesg for a few different attempts, they seem to vary:
10:37fdobridge_: <redsheep>
10:37fdobridge_: <redsheep> ```[ 1225.131093] nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:160 type:109 scope:1 part:233
10:37fdobridge_: <redsheep> [ 1225.131100] nouveau 0000:01:00.0: fifo:c00000:0014:00a0:[bg3_dx11.exe[17975]] errored - disabling channel
10:37fdobridge_: <redsheep> [ 1225.131107] nouveau 0000:01:00.0: bg3_dx11.exe[17975]: channel 160 killed!
10:37fdobridge_: <redsheep> [ 1225.131237] nouveau 0000:01:00.0: bg3_dx11.exe[17975]: error fencing pushbuf: -19
10:37fdobridge_: <redsheep> [ 1393.425630] nouveau 0000:01:00.0: Using 47-bit DMA addresses
10:37fdobridge_: <redsheep> [ 1420.654404] nouveau 0000:01:00.0: bg3_dx11.exe[18836]: job timeout, channel 152 killed!
10:37fdobridge_: <redsheep> [ 1422.052276] nouveau 0000:01:00.0: gsp: mmu fault queued
10:37fdobridge_: <redsheep> [ 1422.053636] nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:152 type:31 scope:1 part:233
10:37fdobridge_: <redsheep> [ 1422.053640] nouveau 0000:01:00.0: fifo:c00000:0013:0098:[bg3_dx11.exe[18836]] errored - disabling channel
10:37fdobridge_: <redsheep> [ 1422.053645] nouveau 0000:01:00.0: bg3_dx11.exe[18836]: channel 152 killed!
10:37fdobridge_: <redsheep> [ 1497.243437] nouveau 0000:01:00.0: Xorg[1002]: nv50cal_space: -16
10:37fdobridge_: <redsheep> [ 1501.736070] nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:120 type:109 scope:1 part:233
10:37fdobridge_: <redsheep> [ 1501.736078] nouveau 0000:01:00.0: fifo:c00000:000f:0078:[bg3_dx11.exe[20177]] errored - disabling channel
10:37fdobridge_: <redsheep> [ 1501.736085] nouveau 0000:01:00.0: bg3_dx11.exe[20177]: channel 120 killed!
10:37fdobridge_: <redsheep> ```
10:38fdobridge_: <redsheep> To be clear, so far I can only get that to happen by actually launching the game, not my trace.
10:42fdobridge_: <redsheep> I had hoped that maybe I could use protontricks to work around the issue and run apitrace in the same prefix as bg3 but my protontricks seems to be busted right now
10:44fdobridge_: <redsheep> montjoie: Performance is still expected to be awful on GP108, that chip doesn't have gsp firmware so you're stuck at the super low boot clocks.
10:47fdobridge_: <redsheep> Even on hardware with gsp firmware there's still a ways to go. Some games run great, but they're generally cpu bound, and some games that work are still super slow such as The Witness, which still only yields 20-40 fps for me on a 4090.
10:48fdobridge_: <redsheep> While running at 720p and no MSAA
11:01montjoie: redsheep any way to enhance this situation ?
11:06EisNerd: seems that enabling vulkan support for nouveau changes sth generally, even when not setting the env vars
11:10EisNerd: as heroic does not load without enabling nvk on my rve6, with setting the env vars
11:11EisNerd: shouldn't not setting the NVK_I_WANT_A_BROKEN_VULKAN_DRIVER=1 cause default behaviour of mesa?
11:51fdobridge_: <makinbacon21> Well yea that's what I was doing, but some arch specific functions seemed missing from ampere cuz of GSP stuff, but yea I'm just trying with a fork of gp10b
11:52fdobridge_: <makinbacon21> It's annoying cuz t194 would be a better reference prolly but there is 0 stuff for that either cuz nvidia stopped caring abt oss graphics lol
11:53fdobridge_: <tom3026> seems they stopped caring about their linux drivers completly, ~2 months between driver releases and ive had broken vulkan extensions and Xid hard locks for soon 2 ½ year, they have replied on their forum posts but its reaching a point where the HW i bought is becoming so old its time to soon replace it lol
11:55fdobridge_: <tom3026> so my last hope is on @gfxstrand 😄 ❤️
11:56fdobridge_: <tom3026> and well the rest of nouveau devs heh
11:59fdobridge_: <makinbacon21> their linux drivers are fine these days, I just want proper Android userspace lol
12:00fdobridge_: <makinbacon21> and unless someone dumps drive Orin running aaos I think I'm out of luck for using their proprietary stuff
15:30fdobridge_: <gfxstrand> Ugh...
15:32fdobridge_: <redsheep> It's probably just something wrong with the wine setup I'm running it on
15:35fdobridge_: <gfxstrand> Could be. It's been a while since I've set up DXVK manually.
15:37EisNerd: you where talking about intel bug, some days ago, which one is it, I have some text not rendered, I can mark it and copy it, but it is not visible
15:38fdobridge_: <redsheep> I didn't try too hard to fix it. I'm not familiar enough with wine to do it all manually so without working protontricks I'm at the mercy or whatever bottles or lutris hand me, which both don't work and don't log anything useful.
15:40EisNerd: fdobridge_: heroic-launcher is quite good, but has display issues, on my end, I can't nail down
15:46EisNerd: fdobridge_: I was thinking about this yesterday also, and may nvidia has some contracts conflicting with opensource, so they can't internally work on such (clean room approach). So they provided the docs, after having them checked by legal and tried to get some external people for this (collabora)
15:48EisNerd: and changed a there design a bit, to get things easier (gsp)
16:52cwabbott: gfxstrand: btw, if you're implementing ESO you'll probably need the common commits from https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25680
16:53cwabbott: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25680/diffs?commit_id=1dea6fd97cf0f1ac51c01087f79e3622bf9750e4 and https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25680/diffs?commit_id=9e923b5b60ca0f4cebcda87680190a94930509fb
16:54gfxstrand: cwabbott: Yeah, I see them. That's annoying but also fine.
16:54gfxstrand: I'll pull them into the MR.
20:14Lyude: Looking into pulling all of the platform driver stuff into the nova branch today btw
20:29Lyude: Oh - airlied and dakr - I'm assuming we probably just cherry-pick work individually from asahi's branch so I'll start by doing that, but let me know if there's a different way I should be pulling this into the nova tree
21:34EisNerd: stupid question, especially in this channel, but has anyone had a look at the upcoming ryzen 7 8700g, will it support vulkan fully? So I would have full dxvk and vkd3d support?
21:35DodoGTA: EisNerd: I think even the ancient HD 7770 supports DXVK with the amdgpu KMD
21:37EisNerd: I would also expect as cd red states it should run there current tripple a title in fhd
21:38EisNerd: I think about buying some stationary system instead of a steam deck
22:05fdobridge_: <orowith2os> DodoGTA: isn't it missing VK 1.3, so only specific versions of DXVK?
22:12EisNerd: fdobridge_: this is where I'm unsure regarding the 8700g, as some specs stated vk 1.0
22:13EisNerd: but my rve6 also provides vk1 and it was never speced to do any vk
22:47Lyude: https://gitlab.freedesktop.org/lyudess/linux/-/commits/rvkms in case anyone's curious, about to start hooking up the platform stuff now that i've pulled it in
23:13fdobridge_: <redsheep> That certainly comes with an interesting warning
23:14Lyude: i don't make the rules i just enforce them
23:15fdobridge_: <redsheep> How much of the asahi stuff do you expect will carry over?
23:16Lyude: A good bit of it but I'm not totally sure yet! I'm probably going to find out as I keep going. the thing is asahi does't have their DRM driver in rust, so there's not many bindings for us to pull in from there quite yet
23:16Lyude: sorry
23:16Lyude: *modesetting driver
23:16Lyude: their DRM driver is in rust :p
23:17Lyude: a lot of the modesetting bits I've gotta come up with myself
23:17fdobridge_: <redsheep> So you're going for even more rust then, interesting
23:20fdobridge_: <redsheep> Well hopefully that new modesetting will be able to straighten out all the stuff the current one doesn't support
23:20Lyude: yep! and yeah - it definitely will :)
23:20fdobridge_: <redsheep> And one would hope rust can help squash some of these weird nigh impossible to replicate crashes
23:21Lyude: I think we'll hopefully just have less of those in general. at least in GSP world there's a lot more we can do in terms of debugging
23:22fdobridge_: <redsheep> Really? I'd expect the opposite, why does gsp help?
23:23Lyude: I mean I'm sure there might be a few areas it's more difficult, but for the most part GSP has quite a lot of logs for debugging that you can send to nvidia and actually get decoded to something coherent. that, and we have the source for their driver now so we have an example implementation of pretty much all of GSP's functionality
23:23Lyude: As well: a lot of the more difficult issues I remember us hitting in the past are all in areas I think GSP handles now?
23:24fdobridge_: <redsheep> It certainly seems to abstract away quite a lot from what I've seen poking around openRM
23:24Lyude: (e.g. the multiple opengl contexts breaking on some older chipsets due to hardware race conditions, at least that's a big one that comes to mind that we probably don't have to deal with now)
23:28fdobridge_: <redsheep> Let me know if you ever want me to test it, particularly modesetting stuff. I'm not overly savvy on the kernel but I know monitors
23:28Lyude: of course :)
23:28fdobridge_: <redsheep> And I've got a few interesting test cases, and probably some new ones soon
23:45Lyude: it's probably going to be quite a while before we get that far Z
23:45Lyude: or, well, however long it takes for nova to get far enough to be ready for the modesetting driver