00:02 fdobridge_: <r​edsheep> Where are the kernel patches I would need to play around with !26622 for the memory types that works around this? I would like to see if I can find other zink fails.
00:28 fdobridge_: <m​henning> Are you sure you're rebuilding zink with that patch? It's a zink patch so building nvk is not enough
00:32 fdobridge_: <r​edsheep> I am just using a modified pkgbuild on the aur package vulkan-nouveau-git, it's how I have been doing all of my testing. As far as I know that is rebuilding all of mesa every time I do makepkg on it, but I don't really know. This is all pretty far outside of my wheelhouse.
00:32 fdobridge_: <r​edsheep> I stripped out the parts that are patching it and pointed it at zmike/test
00:33 fdobridge_: <m​henning> No, that will only install a new nvk.
00:34 fdobridge_: <r​edsheep> I see. I did not realize that was possible, I will see if I can make it build zink.
00:36 fdobridge_: <m​henning> If you want to install all of mesa from git, you might be able to point the mesa-git package at that branch
00:37 fdobridge_: <z​mike.> Just use the MR branch, you don't need anything else I'm testing
00:37 fdobridge_: <z​mike.> I only meant I didn't have any wild nvk patches
00:37 fdobridge_: <m​henning> (I typically test with a devenv instead of installing globally though)
00:38 fdobridge_: <r​edsheep> Yeah I think I should probably figure out how to do that, I am probably at the limits of testing all of this the dumb way.
00:45 fdobridge_: <z​mike.> meson . build -Dgallium-drivers=zink
01:10 fdobridge_: <z​mike.> @gfxstrand if only this MR had come up last week you might have been able to read it
01:20 fdobridge_: <g​fxstrand> 😅
01:27 fdobridge_: <r​edsheep> Yep now that I am actually building zink supertuxkart renders beautifully
01:31 fdobridge_: <r​edsheep> That also gets heaven *nearly* working as intended on the extreme preset
01:31 fdobridge_: <z​mike.> LET THE TESTING BEGIN
01:32 fdobridge_: <r​inlovesyou> after building you just `meson devenv -C builddir ./someGLAPP`
01:32 fdobridge_: <r​edsheep> Aside from the crazy blue flashing every few frames it looks great
01:32 fdobridge_: <r​inlovesyou> after building you just `meson devenv -C builddir ./someGLApp` (edited)
01:33 fdobridge_: <r​inlovesyou> https://docs.mesa3d.org/install.html#running-against-a-local-build-easy-way
01:33 fdobridge_: <r​inlovesyou>
01:33 fdobridge_: <r​inlovesyou> :3
01:35 fdobridge_: <r​edsheep> Ok I have singled out the cause of the flashing in heaven, it doesn't happen with MSAA off
01:35 fdobridge_: <r​edsheep> Every single other setting is working great except for MSAA
01:35 fdobridge_: <g​fxstrand> That's plausible
01:38 fdobridge_: <g​fxstrand> Care to give this a try?
01:38 fdobridge_: <g​fxstrand> https://cdn.discordapp.com/attachments/1034184951790305330/1209675506744561714/0001-nvk-Invalidate-the-texture-cache-before-MSAA-resolve.patch?ex=65e7c91b&is=65d5541b&hm=378019778a59b01e843e7d9f37a0077e35d057807daa17168c6c718f51b49d5f&
01:40 fdobridge_: <r​edsheep> Certainly, one moment
01:48 fdobridge_: <r​edsheep> I'm not doing that devenv thing yet and I caught the keyboard breaking on reboot disease
01:49 fdobridge_: <r​edsheep> Hmm nope, still flashing like crazy
01:57 fdobridge_: <r​edsheep> Doesn't look like supertuxkart has an msaa setting or I would try it there, let me see if I can find any other good test cases to narrow it down
02:02 fdobridge_: <a​irlied> stk running for me on zink on ampere
02:04 fdobridge_: <r​edsheep> MSAA is working fine in xonotic with the ultra preset
02:06 fdobridge_: <g​fxstrand> Okay, it's something else then. I'm still pretty sure that patch is needed so I'll CTS and merge it.
02:07 fdobridge_: <g​fxstrand> Okay, it's something else then. I'm still pretty sure that patch is needed, though, so I'll CTS and merge it. (edited)
02:07 fdobridge_: <g​fxstrand> With Heaven, are you running with DXVK or GL+Zink?
02:08 fdobridge_: <r​edsheep> To my eye the flashing in heaven looks a little bit like the blue background coming through intermittently, but I am not 100% sure
02:08 fdobridge_: <r​edsheep> Is there a directx version of heaven? I am doing gl+zink
02:08 fdobridge_: <g​fxstrand> Yeah, there is if you pull the Windows build
02:08 fdobridge_: <r​edsheep> Oh. Yeah, no I am doing the native linux build
02:09 fdobridge_: <a​irlied> heaven is also working for me
02:09 fdobridge_: <r​edsheep> Are you able to do msaa?
02:09 fdobridge_: <g​fxstrand> With MSAA?
02:09 fdobridge_: <a​irlied> renders with a bit of a flicker
02:10 fdobridge_: <r​edsheep> Yep, ok so same result on ampere then
02:10 fdobridge_: <a​irlied> like a wierd blue overlay
02:12 fdobridge_: <a​irlied> yeah that patch doesn't help it
02:13 fdobridge_: <r​edsheep> As a sidenote, I am pretty sure the heaven performance is way higher than where it was two months ago, so that's nice.
02:14 fdobridge_: <z​mike.> Maybe because you're not using a super old zink
02:15 fdobridge_: <r​edsheep> When I had been testing in December my mesa-git aur package was at most a month old, so not overly ancient on the zink version
02:26 fdobridge_: <r​edsheep> Hmm the togl path in TF2 still doesn't work for me with zink or without
02:26 fdobridge_: <r​edsheep> The new x64 dxvk path is working great though
02:55 fdobridge_: <r​edsheep> Bioshock infinite is rendering well with zink+nvk, though that doesn't appear to have any msaa.
02:58 fdobridge_: <r​edsheep> Hmm. Tomb raider 2013 is failing to launch.
03:00 fdobridge_: <z​mike.> the launcher scripts for a lot of feral ported games are broken
03:01 fdobridge_: <z​mike.> they need to be updated to add something like `LD_LIBRARY_PATH=libs/x86_64` (or whatever the path in the game dir is)
03:09 fdobridge_: <r​edsheep> That's annoying, yeah I added this to the fifth line of the script and it launches:
03:09 fdobridge_: <r​edsheep>
03:09 fdobridge_: <r​edsheep> ```export LD_LIBRARY_PATH="${GAMEROOT}/lib/i686:${LD_LIBRARY_PATH}"```
03:11 fdobridge_: <z​mike.> yup
03:13 fdobridge_: <r​edsheep> Again no msaa in this, but it is rendering flawlessly. So only known issues I have now are msaa in heaven and tf2 togl, and that one might just be broken for other reasons.
03:16 fdobridge_: <a​irlied> @gfxstrand so I supsect the kepler vram mapping not working might be related to us not giving the kind to the kernel
03:17 fdobridge_: <a​irlied> I wonder do we need a new uapi for mapping
03:23 fdobridge_: <r​edsheep> Nevermind on tf2, I didn't know the native version there also has similar issues loading libraries. Adding LD_PRELOAD=/usr/lib32/libtcmalloc.so.4 gets that one up to scratch, so even togl is functional.
03:23 fdobridge_: <r​edsheep>
03:23 fdobridge_: <r​edsheep> That msaa issue in heaven is the only actual issue left out of 7 different applications tested on zink.
03:30 fdobridge_: <z​mike.> @karolherbst is it time to bridge irc #zink ?
03:36 fdobridge_: <k​arolherbst🐧🦀> there is a #zink?
03:36 fdobridge_: <z​mike.> ...yes?
03:36 fdobridge_: <k​arolherbst🐧🦀> mhh
03:37 fdobridge_: <k​arolherbst🐧🦀> I kinda hate the bridging tho :ferrisUpsideDown: but I also wasn't the one setting any of that up
03:38 fdobridge_: <z​mike.> it seems like there's going to be more zink discussion here in the future, and I don't want the discussion to be entirely disassociated from irc
03:38 fdobridge_: <k​arolherbst🐧🦀> yeah, fair
03:38 fdobridge_: <z​mike.> so the choices are either bridge or I'll probably stop replying to zink questions here since irc is "official"
03:38 fdobridge_: <z​mike.> :grimace:
03:39 fdobridge_: <k​arolherbst🐧🦀> ping me tomorrow as I _should_ go to sleeping, because it's almost 5 am 🥲
03:39 fdobridge_: <z​mike.> k
03:47 fdobridge_: <g​fxstrand> We could potentially just restrict mapped VRAM on Kepler to only linear BOs.
03:48 fdobridge_: <a​irlied> probably makes sense to do that for everyone
03:50 fdobridge_: <g​fxstrand> Not so much... VK_EXT_host_copy_image is a thing.
03:51 fdobridge_: <g​fxstrand> Not so much... VK_EXT_host_image_copy is a thing. (edited)
04:03 fdobridge_: <a​irlied> so one thing I'm not sure about, but since nouveau maps tiled images into the BAR, if there is some detiling done there
04:28 fdobridge_: <g​fxstrand> Possibly but not that I've observed
04:35 fdobridge_: <a​irlied> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27712 some trivial memory leak fixes
04:42 fdobridge_: <g​fxstrand> Rb. Merge it
04:47 fdobridge_: <a​irlied> @gfxstrand I left some comment on the mem types MR
04:47 fdobridge_: <g​fxstrand> Okay, cool. I'll deal with them tomorrow or Thursday
05:44 fdobridge_: <a​irlied> tried that trace crash on ampere gsp, didn't take it out, will plug in a turing
05:51 fdobridge_: <a​irlied> not happening on my 2070 either
06:04 fdobridge_: <p​rop_energy_ball> Does it ever work well?
06:05 fdobridge_: <r​edsheep> Well at the end of the day with the loader workaround it did, yes. At least as well as it does on any driver. It ran pretty smooth.
06:09 fdobridge_: <r​edsheep> Actually I think it stuttered less than the dxvk path, but that probably won't be true once 27024 merges
06:42 fdobridge_: <S​id> welp
06:42 fdobridge_: <S​id> looks like my time with running the -rc kernels is over
06:50 fdobridge_: <S​id> in the infinite wisdom of my cursed mind
06:50 fdobridge_: <S​id> I'm doing something that absolutely needs me to use ZFS
12:29 fdobridge_: <k​arolherbst🐧🦀> it's already the news :ferrisUpsideDown: https://www.phoronix.com/news/Zink-NVK-For-NVIDIA-OpenGL
13:13 fdobridge_: <p​avlo_it_115> How is NOUVEAU_USE_ZINK=1 different from MESA_LOADER_DRIVER_OVERRIDE=zink
13:13 fdobridge_: <p​avlo_it_115> Are there any special fixes and patches in the first version to improve compatibility? Is it just a reduction in the team
13:13 fdobridge_: <p​avlo_it_115> How is NOUVEAU_USE_ZINK=1 different from MESA_LOADER_DRIVER_OVERRIDE=zink
13:13 fdobridge_: <p​avlo_it_115> Are there any special fixes and patches in the first version to improve compatibility? Is it just a reduction in the command (edited)
13:13 fdobridge_: <k​arolherbst🐧🦀> it's better on e.g. prime systems
13:14 fdobridge_: <k​arolherbst🐧🦀> so it only replaces the nouveau driver, not also iris or so
13:31 fdobridge_: <p​avlo_it_115> Am I the only one with opengl 4.5 because of zink?
13:33 fdobridge_: <p​avlo_it_115> My bad
13:33 fdobridge_: <p​avlo_it_115> nvk is not implemented VK_EXT_depth_clamp_zero_one
13:35 fdobridge_: <p​avlo_it_115> The Galium driver is faster than zink
13:35 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209855969966686289/b9551155e670e683.png?ex=65e8712d&is=65d5fc2d&hm=40e857a49df6c2eeae871072276c3841367b540b82dcec1f9586a784bfad6ad2&
13:35 fdobridge_: <p​avlo_it_115> https://tenor.com/view/jetstream-sam-my-beloved-gif-22029076
13:35 fdobridge_: <p​avlo_it_115> Galium driver is faster than zink (edited)
13:35 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209855969966686289/b9551155e670e683.png?ex=65e8712d&is=65d5fc2d&hm=40e857a49df6c2eeae871072276c3841367b540b82dcec1f9586a784bfad6ad2&
13:40 fdobridge_: <k​arolherbst🐧🦀> that's CPU overhead tho
13:43 fdobridge_: <p​avlo_it_115> And do nothing about it?
13:43 fdobridge_: <k​arolherbst🐧🦀> correct
13:44 fdobridge_: <k​arolherbst🐧🦀> well..
13:44 fdobridge_: <k​arolherbst🐧🦀> you could also run a CPU profiler and check out where the hot path is
13:44 fdobridge_: <p​avlo_it_115> Unigine_Heaven-4.0 crash
13:44 fdobridge_: <p​avlo_it_115>
13:44 fdobridge_: <p​avlo_it_115> ERROR: unknown nir_intrinsic_op terminate
13:44 fdobridge_: <p​avlo_it_115> MESA: error: ZINK: vkCreateGraphicsPipelines failed (VK_ERROR_UNKNOWN)
13:44 fdobridge_: <p​avlo_it_115> Received signal SIGSEGV, invalid memory reference
13:44 fdobridge_: <k​arolherbst🐧🦀> heh
13:44 fdobridge_: <p​avlo_it_115> Unigine_Heaven-4.0 crash : (
13:44 fdobridge_: <p​avlo_it_115>
13:44 fdobridge_: <p​avlo_it_115> ERROR: unknown nir_intrinsic_op terminate
13:44 fdobridge_: <p​avlo_it_115> MESA: error: ZINK: vkCreateGraphicsPipelines failed (VK_ERROR_UNKNOWN)
13:44 fdobridge_: <p​avlo_it_115> Received signal SIGSEGV, invalid memory reference (edited)
13:45 fdobridge_: <!​DodoNVK (she) 🇱🇹> codegen momento
13:47 fdobridge_: <p​avlo_it_115> What tool to use? Maybe "perf" will do?
13:47 fdobridge_: <p​avlo_it_115> or gprof
13:51 fdobridge_: <k​arolherbst🐧🦀> uhhh... the cli tools all are a UX nightmare, I usually use GUI like gprof or kcachegrind
13:51 fdobridge_: <k​arolherbst🐧🦀> ehh
13:51 fdobridge_: <k​arolherbst🐧🦀> hotspot is the KDE one
13:53 fdobridge_: <g​fxstrand> Hrm... Maybe I need a newer kernel? 🤷🏻‍♀️
14:03 fdobridge_: <p​avlo_it_115> So this is the only way to solve the overclocking problem. Somehow people made VP2-VP5 for <NVF0
14:04 fdobridge_: <p​avlo_it_115> sorry for necroposting :-P
14:04 fdobridge_: <p​avlo_it_115> |sorry for necroposting :-P|| (edited)
14:04 fdobridge_: <p​avlo_it_115> ||sorry for necroposting :-P|| (edited)
14:05 fdobridge_: <k​arolherbst🐧🦀> I mean.. sure, one _could_ do that, but it's also a nightmare to support
14:05 fdobridge_: <p​avlo_it_115> well, at least in some condition...
14:07 fdobridge_: <p​avlo_it_115> There are no other variant. Nvidia will never provide normal firmware for this
14:08 fdobridge_: <p​avlo_it_115> There are no other variant. Nvidia will never provide normal firmware for this.. (edited)
16:23 fdobridge_: <z​mike.> KHR-GL46.buffer_storage.map_persistent_dispatch
17:57 fdobridge_: <g​fxstrand> Yeah, I should probably look at what we need to do to finish off Maxwell and drop codegen back to Kepler.
17:57 fdobridge_: <g​fxstrand> I think we're getting close
18:22 soreau: would it be possible to use zink with the proprietary nvidia vulkan driver or any other !mesa vulkan driver?
18:28 fdobridge_: <a​irlied> It works with NVIDIA already
18:32 soreau: good to know, thanks
18:35 soreau: btw, nice work on all of this development.. it's impressive
18:42 fdobridge_: <a​irlied> Can't think of anything that would fix it but who knows
18:45 fdobridge_: <m​ohamexiety> https://www.supergoodcode.com/woof/
19:09 fdobridge_: <a​irlied> @zmike must be the memleak fix from yesterday causing those crashes
19:14 fdobridge_: <z​mike.> https://cdn.discordapp.com/attachments/1034184951790305330/1209941167416938597/image.png?ex=65e8c085&is=65d64b85&hm=6e7a0192219636c24771be89f4623759c90193e221376251506701152bf5ee5c&
19:26 fdobridge_: <z​mike.> think that should do it for triaging the regressions I found since I originally made the cts baseline
19:26 fdobridge_: <z​mike.> updating piglit now
19:29 fdobridge_: <p​rop_energy_ball> @zmike. I think we should also enable Zink by default for a7xx given it's the only choice there too, right?
19:30 fdobridge_: <z​mike.> it works automatically there
19:30 fdobridge_: <p​rop_energy_ball> Is not for me atm...
19:30 fdobridge_: <p​rop_energy_ball> Am I missing a patch in my tree?
19:30 fdobridge_: <p​rop_energy_ball> Maybe zink didn't end up building which would be funny...
19:30 fdobridge_: <z​mike.> the only reason changes were needed for nouveau is because the GL driver supports those gpus
19:31 fdobridge_: <p​rop_energy_ball> Wonder why that isn't working for me then... I'll prod at it
19:31 fdobridge_: <z​mike.> mesa has had implicit loading of zink since last...october?
19:31 fdobridge_: <z​mike.> if no other hw driver loads, zink is loaded
19:31 fdobridge_: <z​mike.> or should
19:32 fdobridge_: <p​rop_energy_ball> Oh, it looks like Freedreno tries to init, fails, then zink never gets a chance to load because FORTIFY_SOURCE detected some bullshit and everything died 🐸
19:32 fdobridge_: <p​rop_energy_ball> `pthread_mutex_destroy called on a destroyed mutex`
19:32 fdobridge_: <p​rop_energy_ball> nice
19:33 fdobridge_: <p​rop_energy_ball> anyway #ID:1034184951790305330 🐸
19:39 fdobridge_: <a​irlied> @zmike nope not seeing those crashes here at all, and the leak fix doesn't look like it could be broken
19:40 fdobridge_: <a​irlied> if you run the test standalone does it die?
19:41 fdobridge_: <!​DodoNVK (she) 🇱🇹> Technically Adreno is AMD/ATI-based so you're talking about a NVIDIA competitor's GPU 🐸
19:45 fdobridge_: <z​mike.> the one in the title crashes solo
19:48 fdobridge_: <a​irlied> wierd, not dying for me, on current origin/main? can you valgrind it?
19:48 fdobridge_: <z​mike.> on main
19:48 fdobridge_: <z​mike.> valgrind never finished, asan didn't give me a stack
19:49 fdobridge_: <z​mike.> can't change mesa versions atm since I'm doing a piglit run
19:50 fdobridge_: <p​rop_energy_ball> 🐖 🔥 ❗
19:50 fdobridge_: <a​irlied> I see your backtrace has egl in it, I'll check with x11_egl vk-gl-cts
19:51 fdobridge_: <g​fxstrand> I'll give it a try. What kernel are you running?
19:52 fdobridge_: <z​mike.> yeah it's surfaceless config
19:53 fdobridge_: <z​mike.> not sure why that would matter here though
19:57 fdobridge_: <a​irlied> not sure why I don't see any crashes either though 🙂
20:01 fdobridge_: <p​avlo_it_115> It's like I got... IDK how to analyze this
20:01 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209953127046840360/9caaa075311affd0.png?ex=65e8cba9&is=65d656a9&hm=dc7b167cc03e43ec986eab2c6b4717c21c6282076daaaffb158286401a370880&
20:01 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209953127479115836/callgrind.out.11486?ex=65e8cba9&is=65d656a9&hm=cde1f36ab9236bc18395dbb0c454b0fcd449bc1e9ad69eaaaec2b2022ae2f27e&
20:02 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209953179979227187/2878eb7adca8c4be.png?ex=65e8cbb5&is=65d656b5&hm=d01722003293c5b9eafb2797f4d169a4030eee7f7a80ec3c114ad9b35f3a6d29&
20:06 fdobridge_: <z​mike.> looks like a weird amount of time spent in memset?
20:07 fdobridge_: <p​avlo_it_115> this gallium
20:07 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209954575134752890/e2f42cc3a171df85.png?ex=65e8cd02&is=65d65802&hm=eb4b7026cb7d9ecc3e933f87f7fbb1b42f01a090838d6986504d3ee1967eaccb&
20:07 fdobridge_: <p​avlo_it_115> this gallium gl (edited)
20:07 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209954575134752890/e2f42cc3a171df85.png?ex=65e8cd02&is=65d65802&hm=eb4b7026cb7d9ecc3e933f87f7fbb1b42f01a090838d6986504d3ee1967eaccb&
20:09 fdobridge_: <z​mike.> you generally want to sort by the "self" column in the left pane
20:09 fdobridge_: <p​avlo_it_115> They are almost the same compared to gl
20:10 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209955397302095923/callgrind.out.11699?ex=65e8cdc6&is=65d658c6&hm=f42e33d3ea7d875c8c4723d706ea43b0e62a6624284e52168758ad3ae5812aac&
20:12 fdobridge_: <p​avlo_it_115> second column?
20:12 fdobridge_: <z​mike.> yeah
20:13 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209955958801829888/de536d59c22ef8d9.png?ex=65e8ce4c&is=65d6594c&hm=4b88807cd655f504397010d8d4078abb46d65b14e24cf11b820c279bb17b6d61&
20:14 fdobridge_: <p​avlo_it_115> external challenges
20:14 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209956271403438150/5979330a5c60dfe6.png?ex=65e8ce97&is=65d65997&hm=ab4a2b6fa85e725706200caa3f8cb7801c037c6022eca87947cdd36dcec05825&
20:14 fdobridge_: <p​avlo_it_115> external challenges
20:14 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209956444007432263/98ff8b2b502dbfb2.png?ex=65e8cec0&is=65d659c0&hm=e1db4dbd176724954d7f0c3a889bdded31fca8602c0c3cae33abf0bb04641f66&
20:15 fdobridge_: <z​mike.> huh
20:15 fdobridge_: <z​mike.> so I guess the memset in `post_submit` must be heavy at that speed
20:16 fdobridge_: <z​mike.> not surprising I guess since it's so huge
20:18 fdobridge_: <a​irlied> 64k is sizeable
20:20 fdobridge_: <z​mike.> I'll have to take another look at how radeonsi manages the bufferlist to avoid it
20:20 fdobridge_: <z​mike.> once I finish yet another wsi issue :fullheadache:
20:22 fdobridge_: <a​irlied> seems to be exactly the same
20:24 fdobridge_: <z​mike.> and yet the issue is less visible there
20:24 fdobridge_: <!​DodoNVK (she) 🇱🇹> Is it Kopper-related?
20:25 fdobridge_: <z​mike.> it's stupid-related is what it is
20:26 fdobridge_: <p​avlo_it_115> I wonder what to do with it
20:28 fdobridge_: <p​avlo_it_115> I wonder what to do with it...
20:28 fdobridge_: <p​avlo_it_115> Why does memtest index so often? (edited)
20:28 fdobridge_: <p​avlo_it_115> I wonder what to do with it...
20:28 fdobridge_: <p​avlo_it_115> Why does memtest causes so often? (edited)
20:28 fdobridge_: <a​irlied> @zmike maybe you use more batches
20:30 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209960295590658108/bb312b4d05de541e.png?ex=65e8d256&is=65d65d56&hm=c9f7bbf180da4241e353996216819e839d166e2ef399f1d9f52b5ee88d65ca2c&
20:31 fdobridge_: <z​mike.> it's one per frame
20:31 fdobridge_: <z​mike.> also if you're gonna @ someone you gotta use tab or else it doesn't actually ping
20:35 fdobridge_: <r​edsheep> I wonder how accurate the GPU utilization in dxvk overlay is, there are lots of games that feel CPU bound but read high utilization there.
20:35 fdobridge_: <r​edsheep> In the absence of proper data on that from the GPU is there any other reliable way to detect where you're bound?
20:39 fdobridge_: <r​edsheep> The frames go through a fifo queue somewhere right? Maybe recording the average queue length would be informative.
20:40 fdobridge_: <p​avlo_it_115> how do i do it
20:40 fdobridge_: <p​avlo_it_115> How do i do it? (edited)
20:41 fdobridge_: <r​edsheep> Wish I knew, that's really what I'm asking.
20:55 fdobridge_: <r​inlovesyou> We had a little conversation about this recently
20:55 fdobridge_: <r​inlovesyou> Lemme dig
20:56 fdobridge_: <r​inlovesyou> https://discord.com/channels/1033216351990456371/1171505720349446276/1202292260079669259
20:56 fdobridge_: <r​inlovesyou> It's all just an estimate
20:57 fdobridge_: <r​inlovesyou> It's all just an estimate based on how busy the gpu is (edited)
20:59 fdobridge_: <p​avlo_it_115> I wanted to test zink in CS, but it didn't even start =)
20:59 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209967667952357406/Screenshot_20240221_225837.png?ex=65e8d934&is=65d66434&hm=43c8faf977d10b74b55c759cadfd0f08d43b44785b5407d8c5d29f94e7f08cb3&
20:59 fdobridge_: <p​avlo_it_115> I wanted to test zink in CS, but it didn't even start (cs crashed) =) (edited)
20:59 fdobridge_: <p​avlo_it_115> https://cdn.discordapp.com/attachments/1034184951790305330/1209967667952357406/Screenshot_20240221_225837.png?ex=65e8d934&is=65d66434&hm=43c8faf977d10b74b55c759cadfd0f08d43b44785b5407d8c5d29f94e7f08cb3&
21:00 fdobridge_: <p​avlo_it_115> MESA: error: ZINK: vkCreateGraphicsPipelines failed (VK_ERROR_UNKNOWN)
21:06 fdobridge_: <!​DodoNVK (she) 🇱🇹> I tried something even more cursed :triangle_nvk:
21:06 fdobridge_: <!​DodoNVK (she) 🇱🇹> https://cdn.discordapp.com/attachments/1034184951790305330/1209969416368816128/Screenshot_20240221_230416.png?ex=65e8dad5&is=65d665d5&hm=25248cabd6399278cce1cb95500a79b20115e33a10a3ae2ca75fc5aaeb7d637e&
21:08 fdobridge_: <p​avlo_it_115> Maybe next time I will buy AMD
21:08 fdobridge_: <p​avlo_it_115> or risk it and take 1650)
21:17 fdobridge_: <p​ac85> Looks pink
21:17 fdobridge_: <p​ac85> GTA sa vice city
21:19 fdobridge_: <r​edsheep> Since when do gta games have opengl renderers?
21:22 fdobridge_: <r​edsheep> Oh wait, that's probably wined3d
21:24 fdobridge_: <r​edsheep> I suppose zmike did say every single GL game works so that's valid
21:39 fdobridge_: <!​DodoNVK (she) 🇱🇹> GTA 3/Vice City can technically have one