00:09 fdobridge: <r​edsheep> GTX 980 was September 2014, Maxwell 2 is indeed almost 10 years old.
00:10 fdobridge: <s​amantas5855> daaamn
00:10 fdobridge: <s​amantas5855> and pascal 2017?
00:10 fdobridge: <e​sdrastarsis> 2016
00:10 fdobridge: <s​amantas5855> damn
00:11 fdobridge: <s​amantas5855> they are getting old
00:12 fdobridge: <r​edsheep> Both pascal and Maxwell had their biggest cards the year after initial launch
00:12 fdobridge: <r​edsheep> But yeah most of the sales volume is 2014/2016 release date cards
00:13 fdobridge: <e​sdrastarsis> we too
00:17 fdobridge: <r​edsheep> That makes a lot of sense, the remaining feature work is probably going to be tougher and slower progress. You were saying you have a plan to reduce stalling?
00:17 fdobridge: <g​fxstrand> Yup
00:17 fdobridge: <g​fxstrand> IDK if it'll work but I'm gonna try
00:59 fdobridge: <r​inlovesyou> Can't wait
01:00 fdobridge: <r​inlovesyou> Performance in some situations has already drastically improved, but in others it's hitting some other stuff. I have two games that are badly affected so if you need anything from me to figure out what that is let me know
01:01 fdobridge: <r​inlovesyou> Minecraft shaders on high settings going from a top of 20 to a top of 75 is an example of amazing gains made over the last few months in zink & nvk
01:02 fdobridge: <r​inlovesyou> Minecraft shaders on high settings going from a top of 20 to a top of 75 is an example of amazing gains made over the last few months in zink & nvk :MikuLove: (edited)
08:41 fdobridge: <m​agic_rb.> if i managed to properly convince nixpkgs and chaotic-nyx to cooperate i should be compiling faith's modifiers branch now, we'll see when it finishes whether gamescope works
08:43 fdobridge: <m​agic_rb.> huh, freedesktop is also using marge bot, same as the haskell people
08:50 fdobridge: <!​DodoNVK (she) 🇱🇹> I wonder who came up with the idea to use a Simpsons character as the bot name
11:27 fdobridge: <s​elaaaa> well, hi there. i recently decided to embark on the journey of seeing if nvk actually works for my maxwell gpu (a GTX 970). vkcube works, so i wanted to try zink, but it seems to complain:
11:27 fdobridge: <s​elaaaa> ```
11:27 fdobridge: <s​elaaaa> WARNING: Some incorrect rendering might occur because the selected Vulkan device (GeForce GTX 970) doesn't support base Zink requirements: scalarBlockLayout OR EXT_scalar_block_layout
11:27 fdobridge: <s​elaaaa> ```
11:27 fdobridge: <s​elaaaa> and it doesn't like it at all, to the point of crashing even `glxinfo`!
11:27 fdobridge: <s​elaaaa> `Thread 3 "glxinfo:zfq0" received signal SIGSEGV, Segmentation fault.`
11:27 fdobridge: <s​elaaaa> here's a backtrace:
11:27 fdobridge: <s​elaaaa> ```
11:27 fdobridge: <s​elaaaa> (gdb) bt
11:27 fdobridge: <s​elaaaa> #0 0x00007fffebca8fd9 in nvk_queue_submit_drm_nouveau () at ../src/nouveau/vulkan/nvk_queue_drm_nouveau.c:304
11:27 fdobridge: <s​elaaaa> #1 0x00007fffebca9b8d in nvk_queue_submit () at ../src/nouveau/vulkan/nvk_queue.c:295
11:27 fdobridge: <s​elaaaa> #2 0x00007fffebc5829a in vk_queue_submit_final () at ../src/vulkan/runtime/vk_queue.c:378
11:27 fdobridge: <s​elaaaa> #3 0x00007fffebc59114 in vk_queue_submit () at ../src/vulkan/runtime/vk_queue.c:850
11:27 fdobridge: <s​elaaaa> #4 0x00007fffebc5963e in vk_common_QueueSubmit2KHR () at ../src/vulkan/runtime/vk_queue.c:1175
11:27 fdobridge: <s​elaaaa> #5 0x00007fffebc5c78c in vk_common_QueueSubmit () at ../src/vulkan/runtime/vk_synchronization2.c:419
11:27 fdobridge: <s​elaaaa> #6 0x00007ffff64125ce in submit_queue () at ../src/gallium/drivers/zink/zink_batch.c:685
11:27 fdobridge: <s​elaaaa> #7 0x00007ffff5ab30ad in util_queue_thread_func () at ../src/util/u_queue.c:309
11:27 fdobridge: <s​elaaaa> #8 0x00007ffff5aa7a18 in impl_thrd_routine () at ../src/c11/impl/threads_posix.c:67
11:27 fdobridge: <s​elaaaa> #9 0x00007ffff7fb822e in start (p=0x7fffec4a4a38) at src/thread/pthread_create.c:207
11:27 fdobridge: <s​elaaaa> #10 0x00007ffff7fba82f in __clone () at src/thread/x86_64/clone.s:22
11:28 fdobridge: <s​elaaaa> ```
11:28 fdobridge: <s​elaaaa> i'm on alpine stable so i'm still using mesa 23.3.6, with a modified APKBUILD so it doesn't strip the debug symbols and so it actually compiles the vulkan nouveau driver.
11:50 fdobridge: <!​DodoNVK (she) 🇱🇹> You should use the Git version of Mesa for both zink and NVK for the best experience
11:51 fdobridge: <r​edsheep> Also currently NVK is best used with turing cards and up, so if updating still doesn't get yield results that's not too surprising
11:51 fdobridge: <r​edsheep> yield good results*
11:51 fdobridge: <s​elaaaa> yeah, tbh i didn't expect anything to work at all, i was just trying it for fun
11:52 fdobridge: <s​elaaaa> i was a bit worried about that since dependencies and whatnot would be outdated relative to mesa's version
11:52 fdobridge: <s​elaaaa> but i could try it
11:52 fdobridge: <s​elaaaa> i could also switch to alpine edge but.. yeah
12:01 fdobridge: <s​elaaaa> yup, i need to switch to alpine edge since it needs a newer meson.. mm, okay. i'll upgrade and come back again with maybe results
12:02 fdobridge: <m​agic_rb.> i found a weird detail, to launch glxgears on nvk, i need to do `DRI_PRIME=1 NOUVEAU_USE_ZINK=1 gamescope -- bash -c 'DRI_PRIME=0 glxgears' bash` gamescope swaps the gpus
12:07 fdobridge: <!​DodoNVK (she) 🇱🇹> NVK and zink have both received a lot of bug fixes because zmike/gfxstrand are trying to complete OpenGL CTS with no failures/flakes
12:29 fdobridge: <s​elaaaa> after having the equivalent of a near-death experience for computer nerds (my wayland session crashed while i was upgrading my system, and then apk behaved weirdly but i found a way to "resume") i'm back and will be recompiling mesa packages now
12:31 fdobridge: <!​DodoNVK (she) 🇱🇹> @airlied Can this warning be made less severe?: <https://github.com/torvalds/linux/blob/v6.8/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c#L1187> (right now it prints a full kernel trace on my laptop)
13:54 fdobridge: <s​elaaaa> hi agane. the absolute latest version of mesa i'm able to build atm is 24.0.7, the 24.1 release candidates just don't build as there seems to be missing header files, and i can't "just build from git" because alpine's versioning doesn't allow arbitrary versions, it must follow a spec, so i kind of just didn't bother replacing stuff (and because i'd likely need to modify scripts quite a bit, and i already had to just by enabling nvk..)
13:54 fdobridge: <s​elaaaa> in any case. it still crashes. same backtrace, same problem. mm
14:02 fdobridge: <!​DodoNVK (she) 🇱🇹> Can you send the Mesa 24.0.7 backtrace?
14:04 fdobridge: <!​DodoNVK (she) 🇱🇹> (a file is preferred for IRC compatibility)
14:05 fdobridge: <s​elaaaa> oop, fair
14:05 fdobridge: <s​elaaaa> https://cdn.discordapp.com/attachments/1034184951790305330/1239216996449648792/bt.txt?ex=66421e43&is=6640ccc3&hm=d2d2c8fadacfa2b7272512f0d2bea86e4d9a76aed91fd206f9b4e6d3fbddab50&
14:11 fdobridge: <!​DodoNVK (she) 🇱🇹> The line number for the first frame just points to a brace :cursedgears:
14:11 fdobridge: <s​elaaaa> yep!
14:12 fdobridge: <s​elaaaa> it said the same before too
14:13 fdobridge: <!​DodoNVK (she) 🇱🇹> I guess you should just try putting `printf` statements inside various places of that function
14:26 fdobridge: <!​DodoNVK (she) 🇱🇹> It would be funny if apk supported .zip files so dual Android-Alpine Linux packaging could be possible 🔧
14:26 fdobridge: <s​elaaaa> oh god :worry: