00:36 fdobridge: <g​fxstrand> @airlied Eric wants modifiers in my Wednesday if we want it in 24.1. in order for that to happen, you or Karol needs to review the Mesa MR or at least enough to be sure the kernel API is good and we need to merge the kernel patch.
00:36 fdobridge: <a​irlied> yeah I'm going through it at the moment
00:45 fdobridge: <g​fxstrand> Cool, Thanks!
00:56 fdobridge: <a​irlied> just fixing up the conditional for using zink to check the new getparam seems to make things fine in most use cases where I add the use zink by default change on top
00:56 fdobridge: <a​irlied> the copr seems to work my desktop now, with an old kernel and a new kernel with the mod patch
01:00 fdobridge: <a​irlied> though steam is throwing up a kopper error about missing interfaces which suggests I might need more work in that area
01:05 fdobridge: <a​irlied> @zmike. seeing some flickering white on steam app while moving mouse across menus
01:09 fdobridge: <z​mike.> 🤔
01:09 fdobridge: <z​mike.> just on nvk?
01:10 fdobridge: <a​irlied> not sure I've got an alternate test box to confirm
01:10 fdobridge: <a​irlied> seems like a sync issue, since it renderers the menu box in white, then the content
01:10 fdobridge: <z​mike.> 🤔
01:10 fdobridge: <z​mike.> I know steam was working fine very recently
01:10 fdobridge: <z​mike.> something to look at tomorrow I guess
01:16 fdobridge: <a​irlied> https://photos.app.goo.gl/R9noKEUwd75GsS64A
01:56 fdobridge: <M​isyl with Max-Q Design> Your video will be ready soon
01:58 fdobridge: <a​irlied> discord must be doing something
03:07 fdobridge: <a​irlied> bleh still some race in the kernel now, seeing an mmu fault after runpm
03:13 fdobridge: <m​henning> @airlied might be https://gitlab.freedesktop.org/drm/nouveau/-/issues/346
03:15 fdobridge: <a​irlied> ah yes that's the one, just thought fixing the evict stuff fixed that as well
04:28 fdobridge: <a​irlied> @skeggsb9778 cleaned up the dma code and got it working I think
04:29 fdobridge: <a​irlied> but I've got a bunch of further cleanups I think I can do, so I might play around a bit more
05:00 fdobridge: <a​irlied> bleh dma_alloc_contiguous doesn't work on x86 if you want single pages yet
06:34 fdobridge: <a​irlied> bleh dma_alloc_noncontiguous doesn't work on x86 if you want single pages yet (edited)
06:48 airlied: Lyude: sent out a newer fix to the BUG_ON , not sure it's fully correct either, but seems to work here
08:48 fdobridge: <r​edsheep> That video still is not viewable, but I have also seen flicker in steam and it's similar to the flicker I get in various of kde apps.
08:49 fdobridge: <r​edsheep> Given the feedback about correct usage of gamescope with mangohud I am getting a new screenshot so the blig post won't show off using it wrong
08:49 fdobridge: <r​edsheep> Given the feedback about correct usage of gamescope with mangohud I am getting a new screenshot so the blog post won't show off using it wrong (edited)
09:01 fdobridge: <r​edsheep> For whatever reason `gamescope --mangoapp -- %command%` did not result in mangohud rendering so here's `gamescope --mangoapp -- %command%`
09:01 fdobridge: <r​edsheep> https://cdn.discordapp.com/attachments/1034184951790305330/1239502838783017100/image.png?ex=66432879&is=6641d6f9&hm=0b7ec501c5bce62f64dd48fe34530490407fd5d2ea0e3b51e82ad61d30ae80f9&
09:03 fdobridge: <!​DodoNVK (she) 🇱🇹> These commands look the same to me
09:04 fdobridge: <r​edsheep> *sigh* Yeah the actual command used is the one in the screenshot, `gamescope -- mangohud %command%` I just pasted wrong
09:05 fdobridge: <r​edsheep> Just working from Joshie's feedback here: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24795
09:11 fdobridge: <k​arolherbst🐧🦀> does anybody know what games/applications heavily rely on glMultiDraw* being low CPU overhead? @zmike. maybe?
09:13 fdobridge: <r​edsheep> Hmm I am trying to remember if there are any notable RTS games that are native GL. Those would be a great candidate, I think
09:15 fdobridge: <d​jdeath3483> like indirects? gravitymark uses a bunch of 10k+ indirect at 200k asteroids
09:15 fdobridge: <d​jdeath3483> it's got gl/vk backends
09:15 fdobridge: <k​arolherbst🐧🦀> Civilization doesn't use it :ferrisUpsideDown:
09:15 fdobridge: <k​arolherbst🐧🦀> at least 5 doesn't
09:16 fdobridge: <k​arolherbst🐧🦀> which.. I think explains the bad performance 😄
09:16 fdobridge: <k​arolherbst🐧🦀> not indirects, glMultiDraw*, it's just batched draws
09:17 fdobridge: <k​arolherbst🐧🦀> like.. instead of calling draw 100 times with the same state, you call glMultiDraw* with 100 draws
09:17 fdobridge: <d​jdeath3483> hmmm okay, I don't know then
09:32 fdobridge: <S​id> I wonder if Stellaris does
09:32 fdobridge: <S​id> I should test that save file I got from my friend
09:33 fdobridge: <m​agic_rb.> I think unity can do it, whether the devs could be bothered to use multidraw, ive no idea
09:34 fdobridge: <m​agic_rb.> But stellaris is mostly cpu bound afaik so they might have
09:35 fdobridge: <k​arolherbst🐧🦀> well.. so is civilization 5
09:35 fdobridge: <k​arolherbst🐧🦀> and they haven't
09:36 fdobridge: <k​arolherbst🐧🦀> but I don't know what engine they are using
09:36 fdobridge: <S​id> afaik paradox uses their own engine
09:36 fdobridge: <S​id> yup
09:36 fdobridge: <S​id> https://en.m.wikipedia.org/wiki/Paradox_Development_Studio#Game_engines
09:37 fdobridge: <S​id> Clausewitz Engine since 2007
09:37 fdobridge: <k​arolherbst🐧🦀> uhh.. I have no idea if I have any games from them
09:38 fdobridge: <S​id> Europa Universalis
09:38 fdobridge: <k​arolherbst🐧🦀> EU3 and CK2 apparently
09:38 fdobridge: <S​id> Hearts of Iron
09:38 fdobridge: <S​id> yeah there you go
09:39 fdobridge: <k​arolherbst🐧🦀> but like.. if thye don't have a benchmark mode I'm kinda scrwed 😄
09:44 fdobridge: <S​id> they don't
09:44 fdobridge: <S​id> there's a reason I asked my friend for that late game save file :p
12:13 fdobridge: <z​mike.> I think some source engine stuff does
12:13 fdobridge: <z​mike.> unigine benchmarks do
12:13 fdobridge: <z​mike.> oh
12:13 fdobridge: <z​mike.> actual glMultiDraw?
12:13 fdobridge: <z​mike.> no idea
12:14 fdobridge: <z​mike.> DISREGARD
12:14 fdobridge: <k​arolherbst🐧🦀> like whatever ends up calling `draw_vbo` with `num_draws > 1`
12:14 fdobridge: <z​mike.> not necessarily the same
12:14 fdobridge: <z​mike.> that could be multiple draw calls batched together
12:14 fdobridge: <z​mike.> why not just use drawoverhead
12:14 fdobridge: <k​arolherbst🐧🦀> does that happen in a non threaded context?
12:14 fdobridge: <z​mike.> no
12:15 fdobridge: <k​arolherbst🐧🦀> soo nouveau doesn't use threaded context 🙂
12:15 fdobridge: <k​arolherbst🐧🦀> yeah, and the numbers are x60 improvement and all that there, just want to test games being more complex and stuff
12:17 fdobridge: <z​mike.> I don't usually inspect api command streams
12:17 fdobridge: <z​mike.> I just notice sometimes when I get multidraws
12:17 fdobridge: <z​mike.> doom2016 has some
12:19 fdobridge: <k​arolherbst🐧🦀> I see
12:52 fdobridge: <z​mike.> what do I need in my cross file to do a 32bit nvk build
12:52 fdobridge: <z​mike.> for rust
12:58 fdobridge: <k​arolherbst🐧🦀> `rust = '/home/kherbst/.rustup/toolchains/1.59-i686-unknown-linux-gnu/bin/rustc'` or something
12:59 fdobridge: <k​arolherbst🐧🦀> just need to point to the 32 bit toolchain rustc thing
13:01 fdobridge: <z​mike.> fedora doesn't seem to provided i686 packages for rust?
13:01 fdobridge: <z​mike.> fedora doesn't seem to provide i686 packages for rust? (edited)
13:02 fdobridge: <k​arolherbst🐧🦀> yeah dunno
13:02 fdobridge: <k​arolherbst🐧🦀> I don't use distribution provided toolchains
13:03 HdkR: I currently have rusticl and asahi disabled for 32-bit due to the lack of i686 packages and cross compiling problems
13:03 HdkR: In Fedora
13:03 fdobridge: <z​mike.> :grimace:
13:04 fdobridge: <z​mike.> that's not ideal since I apparently have to build this shit to test airlied's issue
13:04 fdobridge: <z​mike.> maybe I can get away with swrast
13:15 fdobridge: <z​mike.> the answer is no
13:24 fdobridge: <r​edsheep> Are you talking about testing the flicker on steam when running with a zink session on the modifiers branch?
13:27 fdobridge: <r​edsheep> He linked to builds he did for fedora that specifically include the modifiers patches and I think they enable zink by default as well https://discord.com/channels/1033216351990456371/1034184951790305330/1238010389271154708
13:31 fdobridge: <z​mike.> hmmm
13:31 fdobridge: <g​fxstrand> Here's my cross file.
13:31 fdobridge: <g​fxstrand> https://cdn.discordapp.com/attachments/1034184951790305330/1239570681868324894/x86.cross?ex=664367a8&is=66421628&hm=e3f812b5fb67bd5ca122948c03b910598821b0288825396dcd5d3a99f674b103&
13:32 fdobridge: <c​langcat> Yes they do and iirc it's limited to one CPU core.
13:33 fdobridge: <c​langcat> Like for paradox games it seems much more likely that your CPU will be the bottleneck long before your GPU is.
13:33 fdobridge: <z​mike.> thx, though that doesn't help with fedora not shipping i686 :/
13:33 fdobridge: <r​edsheep> It would still be better to get a newer build since I don't see one from him that is new enough to have the latest patches
13:35 fdobridge: <g​fxstrand> It's one executable. Are you sure Fedora doesn't package the i686 target?
13:36 fdobridge: <z​mike.> hmmm
13:36 fdobridge: <r​edsheep> Also, do you have a kernel with the vm bind patch needed for NVK modifiers? https://lists.freedesktop.org/archives/nouveau/2024-May/044705.html
13:37 fdobridge: <z​mike.> apparently all I needed was the rust lines
13:42 fdobridge: <z​mike.> ...and a manual rustup
13:42 fdobridge: <z​mike.> https://gitlab.freedesktop.org/mesa/mesa/-/issues/11150 as a tracker
13:45 fdobridge: <z​mike.> ughhhhhhhhhhhh steam can't be apitraced anymore
14:00 fdobridge: <z​mike.> great news, it's not a zink bug
14:04 fdobridge: <z​mike.> conferring with people internally, this is expected behavior for the underlying CEF context
14:04 fdobridge: <z​mike.> so
14:04 fdobridge: <z​mike.> live with it
14:04 HdkR: Quality CEF moment
14:05 fdobridge: <z​mike.> yeah apparently if you're in a composited environment we make the window transparent so that (hopefully) you don't see this
14:05 fdobridge: <z​mike.> but also sometimes we can't enable transparency because CEF does what it wants
14:07 fdobridge: <r​edsheep> Sounds like a different bug from the flicker I had been seeing then, I had hoped they were the same. I don't think Konsole or Kate are CEF based
14:09 fdobridge: <z​mike.> I'm only talking about the steam white rect loading flicker
14:09 fdobridge: <g​fxstrand> What's CEF?
14:09 fdobridge: <r​edsheep> Right
14:09 fdobridge: <r​edsheep> Chromium embedded framework
14:10 fdobridge: <r​edsheep> Browser based app ui
14:10 fdobridge: <g​fxstrand> ah
14:11 fdobridge: <z​mike.> it's teh thing that makes steam work
14:11 fdobridge: <M​isyl with Max-Q Design> or not work depending on who you ask
14:12 fdobridge: <z​mike.> or both at the same time!
14:12 fdobridge: <z​mike.> anyway we're apparently attempting to flatten them down onto the main interface, so at some point this problem should go away on its own
14:12 fdobridge: <M​isyl with Max-Q Design> My imgui UI does that
14:12 fdobridge: <M​isyl with Max-Q Design> Ehe
14:13 fdobridge: <M​isyl with Max-Q Design> Im excited for modifier stuff to be merged
14:23 fdobridge: <r​edsheep> @gfxstrand You will probably want to swap out your screenshot on the MR for this one
14:24 fdobridge: <r​edsheep> So as to not misinform the people with my unknowingly incorrect usage
15:06 OftenTimeConsuming: Hmm, speed-dreams-2 is really the ultimate driver tester - any level that almost reaches 60fps causes a kref push error.
15:24 fdobridge: <c​langcat> Damn if want to test nvk I need to compile Mesa.
15:24 fdobridge: <c​langcat> This is slowly turning into my system just being custom compiled.
15:26 OftenTimeConsuming: I can compile mesa just fine, the main issue with NVK is that it's rusted - you can't compile it without rusting up your toolchain.
15:32 fdobridge: <c​langcat> Ahhh well good to know I need rust installed
15:53 fdobridge: <g​fxstrand> Done. I'd already updated the blog post but not the MR.
15:57 fdobridge: <r​edsheep> Sounds good, I am excited for the read when that goes up
16:16 fdobridge: <!​DodoNVK (she) 🇱🇹> ~~Rewrite Steam in VGUI2~~
16:38 fdobridge: <g​fxstrand> NVK stickers are now up: https://my-store-f6eea7.creator-spring.com/listing/nvk-stickers?product=663
16:50 fdobridge: <z​mike.> smh not even product 666
17:11 fdobridge: <g​fxstrand> And, black stickers if that's more your jam: https://my-store-f6eea7.creator-spring.com/listing/nvk-stickers-black
17:58 fdobridge: <g​fxstrand> @airlied What tree are we planning to land the tiled BO kernel patch through?
18:00 fdobridge: <a​irlied> It'll probably go in via misc-next-fixes so I get it into 6.10rc1 in 2 werks
18:00 airlied: dakr: fyi ^
18:11 fdobridge: <g​fxstrand> Please let me know when it lands.
18:38 dakr: gfxstrand, airlied: Will have a look. I wasn't aware of the patch..
18:38 dakr: I think it did not hit the Nouveau list, probably need to relax the spam filter a bit.
18:39 airlied: not sure if author was subscribed, no idea who does nouveau list moderation if anyone
18:40 airlied: https://patchwork.freedesktop.org/patch/593396/
18:41 airlied: just land that one into drm-misc-next-fixes I think
20:19 fdobridge: <s​keggsb9778> Did you figure out why Lyude's patch broke?
20:45 fdobridge: <a​irlied> Just appears we were getting a kmalloc allocation that wasn't suitable to DMA from which confuses me, using the non coherent API seems to avoid the assert and that problem
20:45 fdobridge: <a​irlied> Not sure if it was just some kmalloc alignment problem, though I did try using a kcache to no avail
22:32 Lyude: gfxstrand: didn't realize you were making stickers!
22:33 Lyude: i should advertise the sticker business my roommate runs more lol
23:07 fdobridge: <g​fxstrand> dakr: Thanks! I just kicked the NVK bits off to marge. 😁