#dri-devel on 2022-06-21 — irc logs at oftc.irclog.whitequark.org

2022-03-22 11:57 ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar

00:01 alyssa has quit [Quit: leaving]

00:06 columbarius has joined #dri-devel

00:08 co1umbarius has quit [Ping timeout: 480 seconds]

00:14 YuGiOhJCJ has quit [Remote host closed the connection]

00:14 YuGiOhJCJ has joined #dri-devel

00:18 adavy has joined #dri-devel

01:04 heat has quit [Ping timeout: 480 seconds]

01:04 <airlied> ajax: 17155 might be one for you

01:05 co1umbarius has joined #dri-devel

01:06 columbarius has quit [Ping timeout: 480 seconds]

01:07 heat has joined #dri-devel

01:22 heat has quit [Ping timeout: 480 seconds]

01:59 tfiga has joined #dri-devel

02:00 rg3igalia has joined #dri-devel

02:02 rib___ has joined #dri-devel

02:08 mclasen has quit [Ping timeout: 480 seconds]

02:18 The_Company has quit []

02:20 YuGiOhJCJ has quit [Remote host closed the connection]

02:21 YuGiOhJCJ has joined #dri-devel

02:41 mhenning has joined #dri-devel

02:41 icecream95 has joined #dri-devel

02:53 Daanct12 has joined #dri-devel

03:09 mhenning has quit [Quit: mhenning]

03:20 saurabhg has joined #dri-devel

03:50 aravind has joined #dri-devel

04:18 lyudess has quit []

04:19 Lyude has joined #dri-devel

04:25 Duke`` has joined #dri-devel

04:40 Daanct12 has quit [Ping timeout: 480 seconds]

04:41 YuGiOhJCJ has quit [Remote host closed the connection]

04:41 YuGiOhJCJ has joined #dri-devel

05:06 tzimmermann has joined #dri-devel

05:08 shankaru has joined #dri-devel

05:09 Daanct12 has joined #dri-devel

05:15 mclasen has joined #dri-devel

05:19 shankaru has quit [Quit: Leaving.]

05:21 shankaru has joined #dri-devel

05:21 itoral has joined #dri-devel

05:37 Duke`` has quit [Ping timeout: 480 seconds]

05:50 devilhorns has joined #dri-devel

05:54 devilhorns has quit []

06:20 mclasen has quit [Ping timeout: 480 seconds]

06:29 MajorBiscuit has joined #dri-devel

06:30 danvet has joined #dri-devel

06:31 saurabhg has quit [Ping timeout: 480 seconds]

06:34 alanc has quit [Remote host closed the connection]

06:39 sul has quit [Ping timeout: 480 seconds]

06:39 MajorBiscuit has quit [Ping timeout: 480 seconds]

06:41 sul has joined #dri-devel

06:43 MajorBiscuit has joined #dri-devel

06:48 frieder has joined #dri-devel

06:57 saurabhg has joined #dri-devel

07:00 alanc has joined #dri-devel

07:00 jkrzyszt has joined #dri-devel

07:01 Lyude has quit [Ping timeout: 480 seconds]

07:02 mvlad has joined #dri-devel

07:08 lygstate has joined #dri-devel

07:10 <lygstate> What's the point of use_elf_tls and USE_ELF_TLS, I've seen them are always defined? can we remove them?

07:12 Daanct12 has quit [Remote host closed the connection]

07:13 Daanct12 has joined #dri-devel

07:13 lemonzest has joined #dri-devel

07:16 <pq> MrCooper, robclark, the only case where LUT and CTM could achieve the same effect is if the CTM is a diagonal matrix, and LUT does not have a limitation that 1.0 input must produce 1.0 output. In any other case there simply is no possible conversion. Those are very severe limitations making the substitution impossible in practise.

07:16 <airlied> jenatali: ^ lygstate question, probably last person to touch it :)

07:17 saurabhg has quit [Ping timeout: 480 seconds]

07:17 <pq> MrCooper, vsyrjala, robclark, a proper way to implement Night Light is to use CTM in light-linear space, meaning you need both DEGAMMA and GAMMA LUTs around it. This will allow changing the white point, which makes the image fool the eye as if not much was changed while blue component is reduced significantly, which is the whole point of Night Light.

07:19 saurabhg has joined #dri-devel

07:20 <pq> vsyrjala, it is ok to expose rgb->ycbcr matrix as CTM (assuming it's a generic matrix) if you also do not make false promises about the DEGAMMA and GAMMA LUT position in the pipeline, e.g. not advertising the CRTC LUTs at all.

07:21 tursulin has joined #dri-devel

07:22 lygstate_ has joined #dri-devel

07:22 lygstate has quit [Read error: Connection reset by peer]

07:25 <pq> Implementing Night Light with a LUT only produces a lower quality effect than with a linear-light CTM. It distorts the perceived colors more.

07:26 frankbinns1 has quit []

07:27 frankbinns has joined #dri-devel

07:30 itoral has quit [Remote host closed the connection]

07:30 itoral has joined #dri-devel

07:35 mi6x3m has joined #dri-devel

07:36 mi6x3m has quit []

07:42 mi6x3m has joined #dri-devel

07:43 <mi6x3m> Hey, I'm experiencing a major performance drop with 22.1.2 compared with 22.1.1

07:43 <mi6x3m> I use crocus and softpipe, did something major changed with the new version?

07:43 <mi6x3m> I also updated libdrm (all 32-bit)

07:46 Lyude has joined #dri-devel

07:54 itoral has quit [Remote host closed the connection]

07:55 itoral has joined #dri-devel

07:55 lynxeye has joined #dri-devel

08:01 kts has joined #dri-devel

08:05 rgallaispou has quit [Read error: Connection reset by peer]

08:18 itoral has quit [Remote host closed the connection]

08:18 itoral has joined #dri-devel

08:33 pixelcluster has joined #dri-devel

08:36 saurabhg has quit [Ping timeout: 480 seconds]

08:42 pcercuei has joined #dri-devel

08:45 saurabhg has joined #dri-devel

08:51 apinheiro has joined #dri-devel

09:04 kts has quit [Ping timeout: 480 seconds]

09:08 kts has joined #dri-devel

09:11 lumag_ has quit [Ping timeout: 480 seconds]

09:14 itoral has quit [Remote host closed the connection]

09:15 itoral has joined #dri-devel

09:17 flacks has quit [Quit: Quitter]

09:18 flacks has joined #dri-devel

09:20 Daanct12 has quit [Ping timeout: 480 seconds]

09:26 <danvet> MrCooper, just cc'ed you on a not-super-public patch, would be great if you can check my thinking

09:29 <MrCooper> sure, haven't seen it yet though

09:30 rkanwal has joined #dri-devel

09:32 srslypascal is now known as Guest2745

09:32 srslypascal has joined #dri-devel

09:32 Guest2745 has quit [Ping timeout: 480 seconds]

09:36 rasterman has joined #dri-devel

09:41 dhromw^ has joined #dri-devel

09:49 mi6x3m has quit [Quit: Leaving]

10:02 camus1 has quit [Remote host closed the connection]

10:03 camus has joined #dri-devel

10:03 sul has quit [Read error: Connection reset by peer]

10:09 flacks has quit [Quit: Quitter]

10:10 flacks has joined #dri-devel

10:15 saurabhg has quit [Read error: Connection reset by peer]

10:17 saurabhg has joined #dri-devel

10:19 YuGiOhJCJ has quit [Quit: YuGiOhJCJ]

10:28 saurabh_1 has joined #dri-devel

10:30 saurabhg has quit [Remote host closed the connection]

10:38 kj has joined #dri-devel

10:39 lumag_ has joined #dri-devel

10:39 icecream95 has quit [Ping timeout: 480 seconds]

10:40 pcercuei has quit [Quit: brb]

10:42 saurabh_1 has quit [Ping timeout: 480 seconds]

10:42 heat has joined #dri-devel

10:44 pcercuei has joined #dri-devel

10:49 shankaru has quit [Quit: Leaving.]

10:50 saurabhg has joined #dri-devel

11:02 kj has quit [Remote host closed the connection]

11:02 saurabhg has quit [Ping timeout: 480 seconds]

11:04 pixelcluster has quit [Remote host closed the connection]

11:09 pixelcluster has joined #dri-devel

11:14 pcercuei has quit [Quit: brb]

11:21 pcercuei has joined #dri-devel

11:21 Company has joined #dri-devel

11:33 sul has joined #dri-devel

11:34 rasterman has quit [Read error: Connection reset by peer]

11:35 rasterman has joined #dri-devel

11:38 sul has quit [Read error: Connection reset by peer]

11:44 itoral has quit [Remote host closed the connection]

11:44 itoral has joined #dri-devel

11:48 sul has joined #dri-devel

11:52 dv_ has joined #dri-devel

11:53 gpiccoli has joined #dri-devel

11:55 aravind has quit []

12:00 sul has quit [Ping timeout: 480 seconds]

12:00 kj has joined #dri-devel

12:01 saurabhg has joined #dri-devel

12:04 <jenatali> @_oftc_lygstate:matrix.org: Yeah, they weren't always defined. Windows used the broken non-TLS path until relatively recently

12:05 sul has joined #dri-devel

12:07 lumag_ has quit [Remote host closed the connection]

12:09 lumag_ has joined #dri-devel

12:15 <jenatali> I think some platforms like Haiku still use the non-TLS path

12:21 sul has quit [Ping timeout: 480 seconds]

12:25 sul has joined #dri-devel

12:26 <karolherbst> haiku doens't have tls?

12:27 lumag_ has quit [Ping timeout: 480 seconds]

12:29 itoral has quit [Remote host closed the connection]

12:34 sul has quit [Read error: Connection reset by peer]

12:40 saurabhg has quit [Ping timeout: 480 seconds]

12:45 kts has quit [Remote host closed the connection]

12:45 kts has joined #dri-devel

12:46 saurabhg has joined #dri-devel

12:50 Akari has quit [Ping timeout: 480 seconds]

12:52 apinheiro has quit [Ping timeout: 480 seconds]

12:56 mclasen has joined #dri-devel

12:58 nchery has joined #dri-devel

13:15 pjakobsson has joined #dri-devel

13:18 frieder has quit [Quit: Leaving]

13:28 <jenatali> Not looking at code yet this morning but that's what I remember from when I tried to fix the buggy non-TLS path before I just added Windows support

13:29 sul has joined #dri-devel

13:31 <jenatali> And speaking of TLS I need someone to sign off on my fix for non-shared-glapi on Windows... It broke a while ago because you can't share TLS across DLL boundaries

13:33 sul has quit [Remote host closed the connection]

13:40 nchery has quit [Remote host closed the connection]

13:41 nchery has joined #dri-devel

13:50 mriesch_ has quit []

13:50 mriesch has joined #dri-devel

14:01 nchery has quit [Ping timeout: 480 seconds]

14:13 dhromw^ has quit [Ping timeout: 480 seconds]

14:16 jkrzyszt has quit [Ping timeout: 480 seconds]

14:20 lygstate has joined #dri-devel

14:21 lygstate_ has quit [Read error: Connection reset by peer]

14:22 Net147_ has joined #dri-devel

14:23 Net147 has quit [Read error: Connection reset by peer]

14:30 shankaru has joined #dri-devel

14:44 <ajax> airlied: oi, egl/x11 doesn't have shm support? you wanna fix that or should i?

14:47 shashanks has joined #dri-devel

15:00 shashanks has quit []

15:02 shashanks has joined #dri-devel

15:17 saurabhg has quit [Ping timeout: 480 seconds]

15:32 <demarchi> tursulin: airlied ok to merge first patch, touching drivers/char/agp/intel-gtt.c, via drm-intel-gt-next? https://patchwork.freedesktop.org/series/105321/

15:34 sul has joined #dri-devel

15:34 sul has quit [Read error: Connection reset by peer]

15:40 ybogdano has joined #dri-devel

15:51 Akari has joined #dri-devel

15:55 <jenatali> karolherbst: Ah looks like Haiku TLS got turned on with https://gitlab.freedesktop.org/mesa/mesa/-/commit/8570a2a280587a1e43ac11ad46ad62dfdd6c7b39

15:55 <jenatali> So, yeah all platforms are using TLS these days, cool, the non-TLS version of glapi and u_current can be deleted

15:55 <anholt_> \o/

15:59 rgallaispou has joined #dri-devel

15:59 tobiasjakobi has joined #dri-devel

16:00 dj-death_ has left #dri-devel [#dri-devel]

16:01 dj-death has joined #dri-devel

16:06 <jani> hwentlan____: 4562236b3bc0 ("drm/amd/dc: Add dc display driver (v2)") added toggling of aconnector->base.override_edid, and I don't really understand why. I'd really like that to be reserved for the debugfs & drm_edid stuff, and having dc modify it makes refactoring of the override code unnecessarily hard

16:07 <jani> hwentlan____: e7b07ceef2a6 ("drm/amd/display: Merge amdgpu_dm_types and amdgpu_dm") moved it to drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

16:11 alyssa has joined #dri-devel

16:13 saurabhg has joined #dri-devel

16:15 MajorBiscuit has quit [Quit: WeeChat 3.5]

16:15 unerlige4 has left #dri-devel [#dri-devel]

16:16 unerlige has joined #dri-devel

16:17 nchery has joined #dri-devel

16:20 jewins has joined #dri-devel

16:21 JohnnyonFlame has quit [Ping timeout: 480 seconds]

16:24 X512 has joined #dri-devel

16:26 <X512> jenatali: Haiku support ELF TLS with __tls_get_addr.

16:26 <X512> Haiku do not support statically linked executables and all executables are linked with `-shared` flag.

16:28 JohnnyonFlame has joined #dri-devel

16:28 <karolherbst> anybody working on isaspec?

16:28 tursulin has quit [Ping timeout: 480 seconds]

16:29 <karolherbst> austriancoder: ohh, seems like you have been working on > 64 bit support in isaspec?

16:29 <karolherbst> looks like I can just start playing around with it and see where it goes

16:32 <austriancoder> karolherbst: jup - should be usable now. I hope to push a WIP MR for etnaviv this year. I got decoding and encoding working but need to look into the a backend IR to finish it.

16:32 kj has quit [Remote host closed the connection]

16:32 <karolherbst> austriancoder: I am currently looking into describing the ISA for nvidia GPUs with it and see how far I get

16:35 gouchi has joined #dri-devel

16:35 <austriancoder> karolherbst: have fun ;)

16:35 <alyssa> Joy... separate_shader isn't getting set for ttn programs

16:36 <karolherbst> thanks

16:37 <alyssa> TGSI_PROPERTY_SEPARABLE_PROGRAM does exist

16:38 <alyssa> but it's absolutely unused

16:39 Akari has quit [Ping timeout: 480 seconds]

16:39 <alyssa> maybe ttn should always set separate_shaders, it might be too conservative but trying to fix up the TGSI tree sounds like a world of pain

16:39 <alyssa> everything in u_simple_shaders assumes a separate shader model, I think

16:41 <alyssa> ditto for HUD

16:41 <alyssa> unsure about nine

16:42 <alyssa> nine looks also separate shader model

16:42 iive has joined #dri-devel

16:42 * alyssa wonders if virglrenderer uses the property

16:42 rkanwal has quit [Ping timeout: 480 seconds]

16:44 <alyssa> morally unused in virglrenderer

16:44 <alyssa> (parsed out and ignored)

16:45 <alyssa> italove: ^^^

16:45 <karolherbst> austriancoder: mhh.. can we have hex support for fields like <gen min=0x170"/> ?

16:46 <karolherbst> mhh I guess python parses it as a string

16:46 <karolherbst> ahh should be easy to fix

16:47 <alyssa> italove: I guess you introduced that in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16308

16:47 <alyssa> I don't understandd the rationale

16:47 <alyssa> "TGSI currently has no concept of a separable shader, and virgl might need in TGSI for performance improvements."

16:47 <alyssa> It's the other way around ... linked shaders can be optimized in ways that separable shaders can't.

16:47 <alyssa> If you don't know if a shader is linked or separable, the default (in Gallium) is separable.

16:48 <alyssa> So I'm not sure what that property was supposed to convey, especially wirht regards to backwards compatibility.

16:48 <alyssa> If there really is an optimization you want in virgl, it should be a TGSI_PROPERTY_LINKED_PROGRAM property, no?

16:50 Duke`` has joined #dri-devel

16:54 <alyssa> I see you wrote a bit about this internally

16:56 <alyssa> I don't quite understand the explanation there, could you explain here? thanks :)

16:56 <karolherbst> austriancoder: do I really have to copy this isa_decode_options struct?

16:57 <karolherbst> or do you have some patches pending, which clean up such common things?

16:58 <anholt_> alyssa: the separable shader prop got added and I couldn't imagine any possible way you could use it.

16:58 <alyssa> anholt_: OK. Would you be ok with a patch to ttn setting nir->info.separate_shaders unconditionally?

16:59 <anholt_> seems reasonable to me.

16:59 <alyssa> (Currently it never sets separate_shader which is breaking panfrost/u_blitter)

16:59 <alyssa> ack, will send that patch, indepdnent of what hapens with the prop

17:00 <austriancoder> karolherbst: at the moment you need to copy it.. I am not in front of a pc right now but I think I do not have a cleanup patch for this in my wip branch

17:01 <karolherbst> okay

17:03 Akari has joined #dri-devel

17:08 maxzor has joined #dri-devel

17:08 saurabhg has quit [Ping timeout: 480 seconds]

17:13 vjaquez has left #dri-devel [#dri-devel]

17:26 cheako has quit [Quit: Connection closed for inactivity]

17:28 sul has joined #dri-devel

17:35 lynxeye has quit [Quit: Leaving.]

17:35 maxzor has quit [Ping timeout: 480 seconds]

17:37 sul has quit [Ping timeout: 480 seconds]

17:41 X512 has quit [Quit: Vision[]: i've been blurred!]

17:50 rkanwal has joined #dri-devel

17:50 maxzor has joined #dri-devel

17:50 sul has joined #dri-devel

17:58 sul has quit [Ping timeout: 480 seconds]

18:07 cheako has joined #dri-devel

18:12 tobiasjakobi has quit []

18:38 mbrost has joined #dri-devel

18:46 <karolherbst> austriancoder: is there a good way to define common "display" bits?

18:46 <karolherbst> ohh wait... I know

18:46 <karolherbst> and no, I guess you have to define it for every instruction, you can just group stuff I suspect

18:49 <austriancoder> karolherbst: maybe use an enum in the xml for common bits?

18:49 <karolherbst> yeah.. something

18:49 <karolherbst> we have this sched block on each instruction where you define stalls, "barriers" and stuff

18:49 <austriancoder> karolherbst: for grouping you can use just a bitset

18:50 <karolherbst> yeah.. I suspect that I will put this sched stuff in a bitset

18:51 <austriancoder> karolherbst: I use an own bitset for an alu src .. something like this: https://www.toptal.com/developers/hastebin/orowirulud.xml

18:52 <karolherbst> okay

18:57 Haaninjo has joined #dri-devel

19:03 jhli has joined #dri-devel

19:08 ngcortes has joined #dri-devel

19:09 jewins has quit [Ping timeout: 480 seconds]

19:10 shankaru has quit [Quit: Leaving.]

19:12 <Lyude> Hey, do we have any way of figuring out why a particular patch hasn't appeared on patchwork?

19:13 <alyssa> Lyude: [insert FUD about email-based patch review here]

19:13 <Lyude> yeah it's a terrible system lmao

19:14 tzimmermann has quit [Quit: Leaving]

19:14 maxzor has quit [Ping timeout: 480 seconds]

19:15 sul has joined #dri-devel

19:16 ella-0_ has joined #dri-devel

19:19 ella-0 has quit [Remote host closed the connection]

19:24 shashank_sharma has joined #dri-devel

19:30 <airlied> ajax: yeah no MIT-SHM in there

19:30 <airlied> I started looking at adding it, but it started hurting my brain too much

19:31 alyssa has left #dri-devel [#dri-devel]

19:31 shashanks has quit [Ping timeout: 480 seconds]

19:32 <ajax> suppose we should fix that, lest we completely nerf llvmpipe perf

19:32 * ajax appends to todo

19:34 LexSfX has joined #dri-devel

19:43 rgallaispou1 has joined #dri-devel

19:44 <airlied> ajax: should we be calling xcb_discard_reply for the put_image reply?

19:44 <airlied> wsi does it, egl doesn't

19:46 rasterman has quit [Quit: Gettin' stinky!]

19:46 <ajax> yes

19:47 rgallaispou has quit [Ping timeout: 480 seconds]

19:47 <ajax> well. i think so anyway. though PutImage doesn't generate a reply...

19:48 <airlied> best to be consistent I suppose

19:58 maxzor has joined #dri-devel

20:00 <airlied> ajax: 17155, has the wsi fix in it as well if you want to just take a quick look

20:03 neonking has joined #dri-devel

20:05 kennylevinsen has joined #dri-devel

20:07 jewins has joined #dri-devel

20:07 rgallaispou1 has quit [Read error: Connection reset by peer]

20:11 rgallaispou has joined #dri-devel

20:13 <karolherbst> uhhhh.. can we drop support for MSVC?

20:13 <karolherbst> we can add it later after it became a real compiler

20:13 apinheiro has joined #dri-devel

20:14 <karolherbst> next time I nack all patches removing _Atomic :P

20:26 <HdkR> karolherbst: Can just switch over to c11 atomics I thought? :)

20:26 <karolherbst> HdkR: heheh.. no

20:26 <karolherbst> apparently people remove _Atomic because MSVC

20:27 <HdkR> huh

20:28 <HdkR> I thought MSVC was C11 at this point. Kind of required to support that keyword at that point

20:29 <karolherbst> it seems like that atomics are optional

20:31 <jannau> yes, there's __STDC_NO_ATOMICS__

20:32 <HdkR> Ah, tested on godbolt. It doesn't support `_Atomic int` but it does support `atomic_int`, that's wacky

20:37 toolchains has quit [Remote host closed the connection]

20:37 <karolherbst> HdkR: wait... seriously?

20:37 <karolherbst> but I suspect Atomic_ is a bit annoying,b ecause I think you can put it on everything

20:37 <karolherbst> uhm.. _Atomic

20:38 <karolherbst> "although unlike other qualifiers, the atomic version of type-name may have a different size, alignment, and object representation" :D

20:38 <karolherbst> but yeah.. guess we should use atomic_* instead

20:40 <HdkR> Sounds reasonable to me

20:41 <jannau> it's still strange as C11 stipulates that 'atomic_int' is a typedef of '_Atomic int'

20:43 <HdkR> yea

20:43 <jannau> I guess it conflicts somehow with the C++11 typedefs in MSVC

20:43 <jannau> looks cursed

20:45 <HdkR> The same could be said for all compiler religions

20:47 <karolherbst> nah.. I am sure only C is that cursed :P

20:48 <Sachiel> C++ursed doesn't read all that well

20:50 <karolherbst> why are you changing C++ instead of just using Rust :P

20:53 gouchi has quit [Remote host closed the connection]

21:01 Duke`` has quit [Ping timeout: 480 seconds]

21:17 mvlad has quit [Remote host closed the connection]

21:18 enunes- has quit []

21:19 enunes has joined #dri-devel

21:22 <anholt_> https://gitlab.freedesktop.org/anholt/deqp-runner/-/merge_requests/41 <-- open to feedback on deqp-runner --fraction and --include behavior.

21:28 <jenatali> _Atomic is on the roadmap apparently, but yeah not there yet

21:35 <karolherbst> austriancoder: can I encode "special values" in isaspec? Like I want a range of bits to be a register index, but the value 255 should be treated as "RZ"

21:56 ngcortes has quit [Remote host closed the connection]

21:56 soreau has quit [Read error: Connection reset by peer]

21:57 soreau has joined #dri-devel

22:09 danvet has quit [Ping timeout: 480 seconds]

22:10 ngcortes has joined #dri-devel

22:26 Haaninjo has quit [Quit: Ex-Chat]

22:35 rkanwal1 has joined #dri-devel

22:35 rkanwal has quit [Read error: Connection reset by peer]

22:38 rasterman has joined #dri-devel

23:02 rkanwal1 has quit [Ping timeout: 480 seconds]

23:03 nchery has quit [Remote host closed the connection]

23:04 enunes has quit [Quit: ZNC - https://znc.in]

23:04 nchery has joined #dri-devel

23:04 apinheiro has quit [Quit: Leaving]

23:06 enunes has joined #dri-devel

23:06 maxzor has quit [Ping timeout: 480 seconds]

23:08 pixelcluster has quit [Quit: Konversation terminated!]

23:34 pcercuei has quit [Quit: dodo]

23:53 FireBurn has joined #dri-devel