06:18 karolherbst: imirkin: uff.. looking into that blender bug caused some weirdo issues, but one issue is a crash inside codegen: "mod u32 %r574 0x00000000 %r572 (0)"
06:18 karolherbst: apparently we don't handle it all that well either
06:18 karolherbst: (I saw the graphicsfuzz bug though)
06:19 imirkin: i think that's literally the same bug
06:19 karolherbst: yeah
06:19 imirkin: just a misplaced assert()
06:19 karolherbst: do you have a fix for it already? otherwise I will dig into it
06:19 imirkin: i do not
06:20 karolherbst: okay
06:20 karolherbst: sadly... is also uses multiple contexts, so this can cause some other fun
06:20 karolherbst: had my gr engine diying hard
06:21 karolherbst: ohh, yeah, that diying engine stuff seems to be fixed by my mt fixes
06:21 karolherbst: I was already wondering why with TGSI = hang and with NIR = that compiler crash
06:22 imirkin: well, 0 mod anything = 0, and 0 divided by anything = 0
06:22 imirkin: so ... not sure why those aren't getting eliminated
06:22 karolherbst: with TGSI I end up with a "mod s32 %r593 0 %r591".. same thing
06:22 karolherbst: yeah...
06:22 imirkin: well, it's signed to begin with
06:22 imirkin: maybe we don't optimize that? we probalby should for ints.
06:22 karolherbst: I am sure it's something stupid
06:23 karolherbst: maybe
06:23 karolherbst: wondering if I should let the intern work on that bug instead :p sounds easy enough
06:24 imirkin: wtvr
06:24 imirkin: it's late
06:24 imirkin:&
07:39 karolherbst: imirkin: btw, here it's not an assert, but a real crash
07:58 karolherbst: and we really don't do any optimization on s == 0 for div or mod
07:59 karolherbst: we might be actually able to do some
07:59 karolherbst: a == 1 should allow us to emit a slct
07:59 karolherbst: + something for s32
08:31 karolherbst: ohh.. it's a crash within the assert, uff
09:57 cosurgi: what debian packages I have to install to have a working GPU acceleration with #nouvesau?
09:57 cosurgi:tries to run a steam game with simplle graphics
10:05 pmoreau: You will need Mesa and possibly the Nouveau DDX (sometimes called xf86-video-nouveau). No idea how those are named in Debian, but they probably have a wiki somewhere with that information.
10:12 cosurgi: thanks. I'm searching now for more info :) oh, btw I have devuan, actually. Hate systemd :)
10:12 cosurgi: reading this https://wiki.archlinux.org/index.php/Nouveau
11:03 cosurgi: pmoreau: how can I know if I have working acceleration?
11:03 cosurgi:tries to find something in Xorg.log
11:04 cosurgi: glxinfo says "direct rendering: Yes"
11:05 cosurgi: [1253386.500] compiled for 1.19.2, module version = 2.6.0
11:05 cosurgi: [1253386.500] ABI class: X.Org Video Driver, version 23.0
11:05 cosurgi: [1253386.500] (II) EXA(0): Driver allocated offscreen pixmaps
11:05 cosurgi: [1253386.500] (II) EXA(0): Driver registered support for the following operations:
11:05 cosurgi: [1253386.500] (II) Solid
11:05 cosurgi: [1253386.500] (II) Copy
11:05 cosurgi: [1253386.500] (II) Composite (RENDER acceleration)
11:05 cosurgi: [1253386.500] (II) UploadToScreen
11:05 cosurgi: [1253386.500] (II) DownloadFromScreen
11:05 cosurgi: [1253386.500] (==) NOUVEAU(0): Backing store enabled
11:05 cosurgi: [1253386.500] (==) NOUVEAU(0): Silken mouse enabled
11:05 cosurgi: [1253386.500] (II) NOUVEAU(0): [XvMC] Associated with Nouveau GeForce 8/9 Textured Video.
11:05 cosurgi: [1253386.500] (II) NOUVEAU(0): [XvMC] Extension initialized.
11:05 cosurgi: [1253386.500] (==) NOUVEAU(0): DPMS enabled
11:05 cosurgi: [1253386.500] (II) NOUVEAU(0): RandR 1.2 enabled, ignore the following RandR disabled message.
11:05 cosurgi: [1253386.501] (--) RandR disabled
11:05 cosurgi: [1253386.503] (II) SELinux: Disabled on system
11:05 cosurgi: Looks like I have this acceleration working??
11:05 cosurgi: but the game with simple graphics (this war of mine) is crazy slow
11:05 cosurgi: like 0.5 fps
11:06 cosurgi: Ahh!
11:06 cosurgi: I see it:
11:06 cosurgi: [1253386.560] (EE) AIGLX: reverting to software rendering
11:06 cosurgi: ok. So it's not eorking.
11:10 pmoreau: cosurgi: Can you please paste the full Xorg.log file to pastebin or some other places?
11:11 cosurgi: pmoreau: https://paste.ubuntu.com/p/6N7NKhJznQ/
11:11 cosurgi: thank you for looking at this.
11:12 cosurgi: I will install libgles1-mesa libglw1-mesa mesa-utils-extra and see if it changes anything.
11:13 pmoreau: “(EE) AIGLX error: Calling driver entry point failed” I don’t think I’ve seen that one before.
11:15 pmoreau: I assume that if you run `glxinfo -B`, you get LLVMPipe or some software rasterizer as vendor?
11:19 barteks2x: did I read what has been said here some time ago right and multithreading will be fixed soon?
11:19 barteks2x: (or already fixed)
11:20 cosurgi: pmoreau: yes:
11:20 cosurgi: direct rendering: Yes
11:20 cosurgi: Device: llvmpipe (LLVM 3.9, 256 bits) (0xffffffff)
11:20 cosurgi: pmoreau: hm, maybe that some binary incompatibility, and I need to recompile mesa.
11:22 pmoreau: That would be weird.
11:23 pmoreau: If you try `LIBGL_DEBUG=verbose glxinfo -B`, what do you get as output? Especially interested in the lines starting with libGL.
11:24 pmoreau: barteks2x: karolherbst has a branch that fixes part of the multithreading issues, but still some issues remain.
11:25 barteks2x: it's so late that soon I will no longer need it fixed...
11:26 cosurgi: pmoreau: https://paste.ubuntu.com/p/cWvwjBvRxG/
11:27 pmoreau: cosurgi: Hum, could you please share the output from dmesg? It looks like there is some kernel issue.
11:29 cosurgi: yes. But is it a problem that I have uptime of 15 days, and dmesg does not go back to reboot?
11:30 cosurgi: the blazer_usb is the UPS driver. It constantly spews some warnings, but UPS connection via USB works and my computer hibernates when there is no electricity..
11:30 cosurgi: hold on pasting
11:30 pmoreau: Ah, that might be problematic as we won’t get the Nouveau initialisation. But maybe there is still some interesting info?
11:30 cosurgi: https://paste.ubuntu.com/p/ZjNw2kn5cr/
11:30 pmoreau: Thanks
11:31 cosurgi: I can filter oyut this blazer tuff..
11:31 cosurgi: I can filter oyut this blazer stuff..
11:32 cosurgi: This is the filtered dmesg: https://paste.ubuntu.com/p/QDCDCNGK5b/
11:35 pmoreau: There is quite a few nouveau errors in there. It would be nice to have one fresh after boot though, to see what are the initial errors
11:38 pmoreau: Oh, you have a Pascal card: make sure you have the firmwares installed. I think that is called firmware-linux on Debian
11:38 cosurgi: yes, I have the firmwares
11:39 cosurgi: I have old syslogs, I'm checking if they have dmesg in them.
11:41 cosurgi: is there some kind ofJul 4 10:54:29 absurd kernel: [ 5.041613] calling nouveau_drm_init+0x0/0x1000 [nouveau] @ 718
11:41 cosurgi: looks like in this syslog I have dmesg boot messages.
11:41 cosurgi: I will extract them for you hold on.
11:46 cosurgi: pmoreau: https://paste.ubuntu.com/p/6KgN4DgsNw/
11:46 cosurgi: pmoreau: that's the syslog from right after reboot. Recovered from ]backup :)
11:47 cosurgi: Is the some kind of verifivation that I have the firmwares?
11:52 cosurgi: I have package firmware-linux-nonfree version 20190114-1
11:52 cosurgi: firmware-misc-nonfree same version 20190114-1
11:53 pmoreau: You would get an error message from Nouveau if it failed to find the firmware, AFAIR
11:53 cosurgi: I think I recompiled them last time when I had some issues with nouveau, which imirkin helped me to fix. The problem wa that some of my screens sometime didn't work.
11:53 pmoreau: Which you didn’t
11:53 cosurgi: It was about 6 months ago.
11:54 cosurgi: He also started to help me with mesa drivers, but at that time I couldn't finish it. Because that was for playing games, and I had work to do :/
11:54 pmoreau: The first error seems to be “nouveau 0000:04:00.0: gr: intr 00000040”
11:55 cosurgi: it's 5 minutes after booting. xservers is definitely up and running by then. I can paste more if you need.
11:57 cosurgi: This messages is also quite frequent in my latest dmesg which I pasted earlier.
11:58 cosurgi: gr: intr 00000040
11:58 pmoreau: I’m looking if there are any fixes in recent kernel version, that could apply for you
11:59 cosurgi: ah. OK. So your diagnosis is that kernel is missing something. And is missing it so much, that not even an error message is produced.
11:59 cosurgi: ?
11:59 pmoreau: Well, you do get an unknown interrupt from the GPU, the intr 00000040.
12:01 cosurgi: ok
12:01 cosurgi: so there is hope :) In latest kernels
12:05 pmoreau: Are you able to try a 5.1 or 5.2 kernel? Maybe using some live CD and running `glxinfo -B` there?
12:08 cosurgi: hm,hm. I'm not sure.
12:08 cosurgi: hold on my libgl1-mesa-dri is version 13.0.6-1+b2
12:09 cosurgi: Maybe I should jsut try installing latest one, like 19.0.2-1 ?
12:10 pmoreau: That might help. Did 13.0 even support Pascal?
12:11 cosurgi: I have no idea
12:11 cosurgi: ok. I would prefer to try that before trying kernel 5.2 on livecd or any other way.
12:13 pmoreau: 13.0 was released in November 2016 and modesetting support was merged in the kernel in Oct. 2016, so it might have been the very first version to support Pascal.
12:13 pmoreau: There have definitely been fixes made in the past 3 years since that release.
12:39 cosurgi: uh-huh. Looks like installing newer libgl1-mesa-dri will be more work. Either I backport a handful of libs, or upgrade devuan ascii to devuan beowulf
12:39 cosurgi: I will have to do this anyway...
12:39 cosurgi: I think I will go in direction of upgrade to beowulf
12:40 cosurgi: debian released stable 2 months ago. So beowulf should be stable pretty soon.
12:40 cosurgi: pmoreau: the libgl1-mesa-dri version on bewulf is 18.3.6-2. Is that promisnig enough?
12:41 cosurgi: Perhaps on bewulf it will be much simpler to backport libgl1-mesa-dri 19.1.2-1
12:45 pmoreau: 18.3 will definitely be better than 13.0
12:47 cosurgi: great. I will do this. But first I need to finish some part of my work. So will get back to this in few hours.
12:54 pmoreau: No problems
20:14 cosurgi: pmoreau: lol, xserver died when I close to finishing my work for today. I guess that speeds up the update :)
20:15 cosurgi: xserver dies once per month for me, because of poor memory managment as imirkin said. When xterms scroll a lot, or chromium refreshes websites a lot suddently there is not enough memory.
20:16 cosurgi: that's because of my resolution 6480x3480
20:18 cosurgi: alternatively I can almost trigger a crash by enabling xcompmgr and waiting few minutes.
20:18 cosurgi: nevermind. I'm updating devuan to testing (which is debian stable 2 months old). So a pretty new system.
23:09 rhyskidd: karolherbst: i dont think i ever ran deus ex md benchmark mode. might have at best only run the shaders in nouveau_shaderdb
23:10 karolherbst: mhhh
23:10 karolherbst: rhyskidd: but you've added the shaders.. no?
23:11 karolherbst: ohhhh
23:11 karolherbst: that was pendingchaos
23:11 karolherbst: ehh
23:11 karolherbst: sorry for that then
23:11 rhyskidd: yeh
23:12 rhyskidd: they should be in there
23:12 rhyskidd: but not from me if i recall correctly
23:12 pendingchaos: IIRC it looked fine when I ran it
23:17 karolherbst: okay
23:17 karolherbst: pendingchaos: I saw some weird light flickering
23:18 karolherbst: normally the light just... you know, flickers how it's supposed to, but for me it was more of a on/off flicker
23:18 karolherbst: looked weird, but didn't investigate