ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
heat has quit [Ping timeout: 480 seconds]
Peste_Bubonica has quit [Quit: Leaving]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #dri-devel
columbarius has joined #dri-devel
co1umbarius has quit [Ping timeout: 480 seconds]
gpoo has joined #dri-devel
gpoo has quit []
Company has quit [Quit: Leaving]
ngcortes has quit [Remote host closed the connection]
saurabhg has joined #dri-devel
technopoirot has quit [Ping timeout: 480 seconds]
<graphitemaster> in glsl is it safe to assume a shared array of T indexed by thread id will be "thread-safe" in the memory barrier case, or is it possible for a write to such an array to be "larger" than T (thus hitting adjacent Ts) and so you need a memory barrier regardless?
<graphitemaster> I cannot find anything in the spec about this
<graphitemaster> Like imagine shared float stuff[N]; stuff[gl_LocalInvocationID] = n, but some stupid compiler decides to generate a 64-bit wide store to the shared variable with like zero extension or something
<graphitemaster> Or like some masked write
<DrNick> OG Alpha style?
<jekstrand> graphitemaster: As long as you stay in-bounds, it should be safe. If you go out-of-bounds, you're in undefined behavior territory anyway.
rsripada has quit [Remote host closed the connection]
Duke`` has joined #dri-devel
TMM has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
TMM has joined #dri-devel
jewins has quit [Read error: Connection reset by peer]
jewins has joined #dri-devel
toolchains has joined #dri-devel
Lucretia has quit []
tzimmermann has joined #dri-devel
Lucretia has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
itoral has joined #dri-devel
sul has quit [Ping timeout: 480 seconds]
srslypascal is now known as Guest3484
sul has joined #dri-devel
srslypascal has joined #dri-devel
Guest3484 has quit [Ping timeout: 480 seconds]
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
toolchains has quit [Remote host closed the connection]
toolchains has joined #dri-devel
srslypascal is now known as Guest3487
srslypascal has joined #dri-devel
Guest3487 has quit [Ping timeout: 480 seconds]
ahajda has joined #dri-devel
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
kts has joined #dri-devel
alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
MajorBiscuit has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
Major_Biscuit has joined #dri-devel
jkrzyszt has joined #dri-devel
<emersion> pq, i agree with you wrt. UINT fourcc formats, thanks for writing that up
camus has quit [Remote host closed the connection]
frieder has joined #dri-devel
camus has joined #dri-devel
MajorBiscuit has quit [Ping timeout: 480 seconds]
lemonzest has joined #dri-devel
maxzor_ has joined #dri-devel
aravind has joined #dri-devel
jewins has quit [Ping timeout: 480 seconds]
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
rasterman has joined #dri-devel
mvlad has joined #dri-devel
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
wvanhauwaert has joined #dri-devel
danvet has joined #dri-devel
lynxeye has joined #dri-devel
<javierm> danvet: yes, but Linus isn't using simpledrm, but simplefb (although isn't clear if is really simplefb or efifb, need to ask him)
<danvet> ah yeah that explains why I was confused :-)
<danvet> and yeah that one is busted
<danvet> uh
<danvet> we don't forcefully nuke the mmaps, so this might be a real bug :-(
<javierm> danvet: in fact, tzimmermann figured it out! He may not even have a driver that uses the firmware-provided framebuffer
<javierm> danvet: we already have fixes in drm-misc-next for that corner case, the sysfb_disable() patches so we might have to backport them as fixes
<danvet> javierm, can you reply on the linus thread with list of sha1 he should cherry-pick for testing?
<javierm> danvet: tzimmermann is typing an email as we speak
<danvet> ah excellent
<mripard> pq: hey, could you review https://lists.freedesktop.org/archives/igt-dev/2022-June/042581.html ? It's the only patch holding back the rest of the patches
tursulin has joined #dri-devel
WinLinux1028 has quit [Read error: No route to host]
kts has quit [Ping timeout: 480 seconds]
LexSfX has quit []
kts has joined #dri-devel
LexSfX has joined #dri-devel
rkanwal has joined #dri-devel
pcercuei has joined #dri-devel
<pq> emersion, thanks, but I feel I came out quite incoherent there. Too little information in the proposal to give a good answer.
<pq> mripard, sure, it's still in my todo list, sorry.
toolchains has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
heat has joined #dri-devel
<mripard> pq: thanks :)
itoral has quit [Remote host closed the connection]
JohnnyonFlame has quit [Ping timeout: 480 seconds]
toolchains has joined #dri-devel
pnowack has joined #dri-devel
pnowack has quit []
itoral has joined #dri-devel
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
toolchains has joined #dri-devel
kj has quit [Remote host closed the connection]
aravind has quit [Ping timeout: 480 seconds]
kj has joined #dri-devel
lplc has quit [Remote host closed the connection]
toolchains has quit [Ping timeout: 480 seconds]
lplc has joined #dri-devel
devilhorns has joined #dri-devel
toolchains has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
aravind has joined #dri-devel
icecream95 has quit [Ping timeout: 480 seconds]
technopoirot has joined #dri-devel
toolchains has joined #dri-devel
toolchains has quit [Remote host closed the connection]
toolchains has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
flacks has quit [Quit: Quitter]
flacks has joined #dri-devel
Company has joined #dri-devel
devilhorns has quit []
toolchains has joined #dri-devel
devilhorns has joined #dri-devel
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
aravind has quit [Ping timeout: 480 seconds]
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
toolchains has quit [Read error: Connection timed out]
kts has quit [Ping timeout: 480 seconds]
shashanks has joined #dri-devel
kts has joined #dri-devel
toolchains has joined #dri-devel
sul has quit [Read error: Connection reset by peer]
itoral has quit [Remote host closed the connection]
itoral has joined #dri-devel
itoral has quit [Remote host closed the connection]
<zmike> jenatali: can you look into this https://gitlab.freedesktop.org/mesa/mesa/-/jobs/24613508
toolchains has quit [Read error: Connection timed out]
rkanwal has quit [Quit: rkanwal]
rkanwal has joined #dri-devel
rkanwal has quit []
rkanwal has joined #dri-devel
<jfalempe> tzimmermann, I was able to reproduce the amdgpu resource clash on my machine.
kts has quit [Quit: Konversation terminated!]
<tzimmermann> jfalempe, oh, great
<jfalempe> And with the 3 commits that you mentioned, they fix the warning.
<tzimmermann> how so?
kts has joined #dri-devel
<tzimmermann> that's even better
<jfalempe> I have unset CONFIG_FB_SIMPLE
<tzimmermann> thanks a lot
<tzimmermann> are you on a mainline kernel?
<jfalempe> I used v5.19-rc4
<javierm> jfalempe: thanks a lot for testing!
<jfalempe> CONFIG_SYSFB=y
<jfalempe> # CONFIG_DRM_SIMPLEDRM is not set
<jfalempe> CONFIG_SYSFB_SIMPLEFB=y
<jfalempe> # CONFIG_FB_SIMPLE is not set
<jfalempe> that's the config I used to reproduce.
ppascher has quit [Ping timeout: 480 seconds]
toolchains has joined #dri-devel
<cwabbott> I assume that commit is what caused it because all the other changes are to freedreno or turnip
<cwabbott> the tests it broke are in the same general area, but I have no idea how it could affect d3d12 because it doesn't use border_color_is_integer
<jenatali> cwabbott: I can take a look later today
<jenatali> cwabbott: zmike: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/gallium/drivers/d3d12/d3d12_context.cpp#L656 - For sampling stencil in D3D you specify the border color as a float. This code would need to convert instead of just memcpying I think
pixelcluster has joined #dri-devel
<cwabbott> jenatali: that struct is a union, so as long as it's just copying bits I guess it should be fine?
<jenatali> cwabbott: No, it's copying integer bits into a destination that's going to interpret it as a float
<cwabbott> also, stencil is just a normal integer border color, so if this is a problem it should be a problem for everything
<cwabbott> jenatali: I think it should hit the lowering path, actually
<jenatali> D3D doesn't support integer texture sampling (yet), so there is no integer border color concept
<jenatali> Except for stencil...
<cwabbott> jenatali: what I don't get is, it already was an integer
<cwabbott> the bits were the same
maxzor_ has quit [Ping timeout: 480 seconds]
<jenatali> Hm? What do you mean it already was an integer?
<cwabbott> stencil border color is an integer
<cwabbott> it just was reinterpreting the integer as float and copying it around (in mesa/st)
<cwabbott> now it copies around the same integer
<jenatali> st_translate_color is getting called with is_integer, and the definition of is_integer is now changed to include stencil sampling where it didn't before
<cwabbott> right
<cwabbott> but if you look at what it's copying around, it's a union
<cwabbott> so the is_integer and !is_integer parts are actually doing the same thing
<cwabbott> (except the cases that replace with a constant)
<cwabbott> the higher layer (mesa itself) should be setting the stencil border color as an integer
<jenatali> I see
<jenatali> Which is missing from the integer side of the function
<cwabbott> that commit also moves it
mclasen has joined #dri-devel
<cwabbott> to the integer side
<jenatali> Oh. Sorry I just woke up, haven't tried it yet :)
<zmike> you mean you don't do border colors at 5:30am!??!!?!??!?
<cwabbott> border colors: not even once
<ccr> "this is your brain." "this is your brain on border colors."
<cwabbott> jenatali: also, I don't think the driver knows whether a sampler is a stencil sampler at shader-compile time
<cwabbott> so the integer sampling lowering should also happen for stencil
<jenatali> Right
<jenatali> I'll see if I can get you an answer before the baby wakes up, otherwise I probably won't be able to get you one until quite later in the day
<cwabbott> haha, ok
rkanwal has quit [Quit: rkanwal]
Lynne has quit [Remote host closed the connection]
shashank_sharma has joined #dri-devel
shashank_s has joined #dri-devel
saurabhg has quit [Remote host closed the connection]
saurabhg has joined #dri-devel
aravind has joined #dri-devel
shashanks has quit [Ping timeout: 480 seconds]
<jenatali> The difference is that now I'm getting border color not-replicated, only in the R channel
shashank_sharma has quit [Ping timeout: 480 seconds]
<jenatali> And for whatever reason D3D samples border color out of G
pjakobsson has quit [Remote host closed the connection]
<jenatali> cwabbott: texobj->StencilSampling is false for me, so st_translate_color gets called with integer format as false but baseFormat as GL_STENCIL_INDEX, sending it down the float path, but it doesn't hit any cases in the switch
<jenatali> So, replicating the GL_STENCIL_INDEX handling to both sides of the integer branch would work. But you probably want to understand why StencilSampling is false?
<cwabbott> jenatali: huh, yeah that sounds broken?
<cwabbott> then again, I don't know much about GL
<cwabbott> but at least is_integer should be true if we're sampling stencil
<jenatali> zmike: ^^
<cwabbott> otherwise vulkan will definitely be broken
<cwabbott> by vulkan I mean zink
<zmike> uhhh
<zmike> which test are we talking about here
<zmike> texwrap formats ugh my ptsd
<cwabbott> I thought GL_STENCIL_INDEX was not a real format, just used for the StencilSampling (sample stencil part of D24S8) case
<cwabbott> but maybe that's wrong?
* jenatali shrugs
<jenatali> My knowledge of GL is limited to what I learn from debugging code that we asked other people to write for us, or me implementing new extensions
<jenatali> And I haven't had to touch stencil sampling yet :)
<zmike> I'm looking at it
<zmike> hmmm I'm seeing is_integer=1 in st_translate_color
<jenatali> I wonder if I'm looking at another MSVC packed enum being signed issue...
<jenatali> Let's see how StencilSampling gets set
<zmike> ah but texobj->StencilSampling is indeed false
<jenatali> Oh weird then how do you get is_integer=1?
<zmike> texobj->_IsIntegerFormat
<jenatali> That's also false for me
<jenatali> Interesting
<zmike> just to make sure, this is piglit/bin/texwrap GL_ARB_texture_stencil8 bordercolor swizzled
<jenatali> Oh I didn't go for the swizzled one
<zmike> isn't that the one that's failing?
<zmike> oh both of them are
<jenatali> Yeah
<zmike> ok so it's fine
<zmike> the line wrapping and my brain
ella-0_ has joined #dri-devel
sul has joined #dri-devel
<cwabbott> does mesa use GL_STENCIL_INDEX internally for stencil formats?
<zmike> hm the samplerview being referenced is PIPE_FORMAT_S8_UINT
<zmike> so it knows it's stencil
<zmike> puzzling
<cwabbott> I guess if that's true it would be expected, but I'd expect IsInteger to be true
<cwabbott> _IsIntegerFormat that is
ella-0 has quit [Read error: Connection reset by peer]
<zmike> ah GL_DEPTH_STENCIL_TEXTURE_MODE isn't used I think
<jenatali> Right, because this is a stencil-only texture, not the stencil part of depth-stencil
<jenatali> So that sounds like a bug, but a different bug
<zmike> I don't know that it's a bug?
<jenatali> Why do you have _IsIntegerFormat true but not me... hm...
<zmike> ah that bug
<zmike> let's find out
nchery has joined #dri-devel
<zmike> texobj.c:709
<zmike> should be it
<jenatali> Right
<zmike> I have GL_UNSIGNED_INT;
<jenatali> Looking...
<jenatali> I have GL_UNSIGNED_NORMALIZED
<zmike> 🤔
<jenatali> Probably because we don't have an S8_UINT format
<jenatali> So we map it to Z24S8
<cwabbott> well, in that case I guess StencilSampling should be true then?
<zmike> huh.
<jenatali> So we need StencilSampling to be true because this is sampling the stencil part of a depth-stencil texture for us
<zmike> so then maybe the check should be (texobj->_IsIntegerFormat || texobj->StencilSampling || texBaseFormat == GL_STENCIL_INDEX)
<zmike> in st_atom_sampler.c
<cwabbott> I imagine you'd want StencilSampling to be true, though
<jenatali> That'd work. Or maybe when testing texture completeness we should set StencilSampling to be true if the base format is GL_STENCIL_INDEX
<zmike> I think maybe you don't want to mess with StencilSampling since that's a GL state
<zmike> this is the only place it should matter anyway, it seems
<jenatali> texobj.h:142 also looks suspect
<jenatali> Agreed those look like the only 2 though
<zmike> hm
<zmike> do you have restrictions on nearest vs linear for stencil sampling?
<zmike> I guess GL does so it doesn't matter if you do
<jenatali> Yeah, it's an integer, we don't support sampling integer formats at all, right now we lower them to loads but in the future we're adding point sampling support
<jenatali> But yeah it's a GL limitation too
<zmike> hm
<zmike> I think it doesn't affect the texobj case
<zmike> that's specifically for stencil sampling of a depth/stencil image
<zmike> so changing that one conditional in st_atom_sampler.c should be sufficient
<jenatali> Ok
<zmike> cwabbott: ^
toolchains has quit [Ping timeout: 480 seconds]
aravind has quit [Ping timeout: 480 seconds]
frieder has quit [Remote host closed the connection]
frieder has joined #dri-devel
kj has quit [Remote host closed the connection]
jewins has joined #dri-devel
<cwabbott> ok, seems like that fixed it
<zmike> \o/
<zmike> cwabbott: marge away!
kj has joined #dri-devel
sul has quit [Read error: Connection reset by peer]
nchery has quit [Ping timeout: 480 seconds]
JohnnyonFlame has joined #dri-devel
sul has joined #dri-devel
nchery has joined #dri-devel
maxzor_ has joined #dri-devel
sul has quit [Read error: Connection reset by peer]
kj has quit [Remote host closed the connection]
maxzor_ has quit [Ping timeout: 480 seconds]
technopoirot has quit [Ping timeout: 480 seconds]
kj has joined #dri-devel
<jenatali> cwabbott: I had to disappear for a bit otherwise I would've said you could add an r-b to that last patch (now that I've been forced to understand it in depth anyway :P)
nchery has quit [Ping timeout: 480 seconds]
tobiasjakobi has joined #dri-devel
chaim has joined #dri-devel
kts has quit [Quit: Konversation terminated!]
nchery has joined #dri-devel
sul has joined #dri-devel
Lynne has joined #dri-devel
toolchains has joined #dri-devel
Duke`` has joined #dri-devel
tzimmermann has quit [Quit: Leaving]
<austriancoder> If someone has some spare time for a quick review: !17121
wvanhauwaert has quit [Ping timeout: 480 seconds]
<jenatali> Looks like the freedreno runners are pretty badly overloaded right now, taking a full hour for a change to common Vulkan code :(
<zmike> robclark playing minecraft again during work hours smh
tursulin has quit [Ping timeout: 480 seconds]
<robclark> zmike: slacker!
<robclark> not sure if there is a way to see what jobs are queued up?
<jenatali> Looks like I got a flake failure too :(
toolchains has quit [Ping timeout: 480 seconds]
tobiasjakobi has quit []
<robclark> jenatali: looks like it is in particular a630 runners that are backed up? Could be related to a thing tomeu is trying to get running (freedreno in vm ci job)?
<jenatali> Ah could be
<jenatali> Guess I'll try again later
flto has quit [Quit: Leaving]
tango_ has quit [Ping timeout: 480 seconds]
<robclark> tomeu, anholt_, I wonder if we could do a separate tag/label thing for a subset of any given group of runners that we could use for WIP/experimental things, so-as to not starve marge?
<robclark> like marge generated pipelines have access to whole set of runners, but other pipelines have just a subset.. or something along those lines?
technopoirot has joined #dri-devel
devilhorns has quit []
<anholt_> could easily do for bare-metal, but harder to do for lava.
<anholt_> (their tags aren't at the gitlab level, and you'd need to coordinate with owners of those labs instead of just going in gitlab admin interface)
<robclark> ahh.. well at least we could do it for a630 runners
flto has joined #dri-devel
ybogdano has joined #dri-devel
<robclark> anholt_, tomeu: hmm, I suppose if the lava runners are doing their own job scheduling, they could still prioritize marge jobs?
<jenatali> It also seems like marge should cancel in-progress jobs when a merge times out and it moves on to a new one, otherwise those old jobs are just stealing time from the new ones
<anholt_> I don't know that lava cares about gitlab-level job cancels, anyway.
<jenatali> Oh then maybe that's not helpful
<anholt_> I mean, it would be really good if it did, but collabora would have to have some daemon outside of gitlab-runner watching and managing that.
<anholt_> (I'm way more concerned about people clicking cancel than marge timeouts, which are quite infrequent)
<jenatali> Fair
ppascher has joined #dri-devel
<krh> is nir_opt_offsets.c useful for global IO? Something like load_global_amd/ir3 seems like could benefit?
<anholt_> seems like
toolchains has joined #dri-devel
<tomeu> anholt_: we have a priority property that we could increase for jobs in marge-triggered pipelines
<tomeu> but right now the gitlab runner for lava will stop taking new jobs if the queue is full
<tomeu> we could do some scheduling in the gitlab runner, but then the time waiting in the queue will count towards the gitlab job timeout
<tomeu> robclark: btw, I only had one long-running job, wonder if more machines were stuck even if the gitlab jobs had already stopped
toolchains has quit [Ping timeout: 480 seconds]
<jekstrand> bbrezillon: Is dozen using the common secondary support these days?
<robclark> not sure, it would be nice to have a timeline view of when individual runners started and finished various jobs
pixelcluster has quit [Quit: Bye-bye!]
jkrzyszt has quit [Ping timeout: 480 seconds]
jewins1 has joined #dri-devel
Haaninjo has joined #dri-devel
jewins has quit [Ping timeout: 480 seconds]
JohnnyonF has joined #dri-devel
sul has quit [Read error: Connection reset by peer]
Johnny has joined #dri-devel
<jenatali> jekstrand: I think so
<jekstrand> Ok, I guess I get to blind-type convert its pipeline layouts too.
<jekstrand> attempting lavapipe at the moment
<jenatali> I can help if you need
toolchains has joined #dri-devel
JohnnyonFlame has quit [Ping timeout: 480 seconds]
JohnnyonF has quit [Ping timeout: 480 seconds]
* jekstrand really wishes he'd payed more attention to VK_KHR_pipeline_libraries
<zmike> ^ - everyone in khronos anytime any extension is released
<jekstrand> I was too trusting
<jekstrand> This descriptor set layout merging is utter nonsense
<jekstrand> And there's absolutely no good reason for it
<zmike> if you think it's bad now you should've seen it before I started hassling people about it
saurabhg has quit [Read error: Connection reset by peer]
frieder has quit [Remote host closed the connection]
technopoirot has quit [Ping timeout: 480 seconds]
<jenatali> Man CI is having a rough time today
<jekstrand> zmike: Ok, talked to tobias and I'm not crazy and it's not as bad as I thought. Lavapipe is just insane.
<jekstrand> I really should have started from that assumption....
Major_Biscuit has quit [Ping timeout: 480 seconds]
<jenatali> Hopefully Dozen is less insane :P
<jekstrand> Dozen doesn't have pipeline libraries yet
<jenatali> True
<zmike> jekstrand: I think I might have been able to simplify the copy code after the craziness was driven out of the spec but instead I just added a null check
<zmike> though also the way lavapipe handles descriptor sets is a little special
toolchains has quit [Read error: Connection timed out]
mbrost has joined #dri-devel
<bbrezillon> jekstrand: yes, it is
<jekstrand> bbrezillon: Ok, I've found a path that doesn't require quite as much effort as my original plan
<bbrezillon> how does secondaries interact with the pipeline library stuff you were mentioning? Or is this 2 different things that popped up in the discussion?
sul has joined #dri-devel
<anholt_> jenatali: yeah, sorry about that. doing some filtering work to reduce the pain
<jekstrand> I'm trying to add a common vk_pipeline_layout implementation
<bbrezillon> that's good news, I guess
<jekstrand> bbrezillon: And we can simplify some of the stuff with secondaries if we can assume every driver that uses the common trace/replay code uses vk_pipeline_layout
<jekstrand> Specifically, we can get rid of those annoying per-diver ref/unref hooks
<bbrezillon> hm, interesting
<bbrezillon> I guess the plan is to keep all the driver specific stuff at the set_layout level
<jekstrand> Yeah.
<bbrezillon> and then just reference set layouts from the pipeline layout
<bbrezillon> too bad I diverged from this approach a while back then...
<jekstrand> And that's where it interacts w/ pipeline libraries. With pipeline libraries, you don't necessarily have the full set of set layouts up-front.
<jekstrand> You just have the ones that a particular shader needs.
<jekstrand> Which is fine for anv/radv
<zmike> ah, I see why you are sobbing in the corner now
<jekstrand> Less fine for panvk but I've got a plan to let panvk keep having its own thing.
<bbrezillon> I guess that just means we can't create the root signature at pipeline-layout creation time
<jekstrand> maybe not?
<jekstrand> I'd have to be far deeper in the details to be sure.
<bbrezillon> that's probably fine
toolchains has joined #dri-devel
<bbrezillon> especially since such pipelines are not supposed to be bound
<zmike> I'd imagine you can't create the root signature until you've got the full pipeline since descriptor sets can be merged at any point
<jenatali> I mean we could create a root signature, but if it gets merged, we'd have to create a different one with the new result
<bbrezillon> yep, there's no point in creating the root signature early if it's not the final one
<jenatali> That sounds okay, those objects are supposed to be relatively cheap, they're just a way to cache driver metadata and don't correspond directly to GPU objects
<bbrezillon> and I guess we have all we need at the pipeline level to create it late
<bbrezillon> all we the the set layouts for is the binding translation step happening in the NIR domain
<bbrezillon> (I mean, in the pipeline-library case)
<bbrezillon> and jekstrand seemed to imply that we get passed all the set-layout shaders care about
saurabhg has joined #dri-devel
<jekstrand> bbrezillon: Yup
<jekstrand> bbrezillon: It's only if you have cross-set stuff that it gets hairy
JohnnyonFlame has joined #dri-devel
JohnnyonFlame has quit []
toolchains has quit [Ping timeout: 480 seconds]
<jenatali> anholt_: One thing I've noticed is that there's 2 a630 jobs that run on every MR, seeming to ignore the rules, the gles_asan and vk_asan ones
<jenatali> Or maybe that's for robclark ?
Johnny has quit [Ping timeout: 480 seconds]
<ajax> anyone happen to know how we do macos ci these days?
<zmike> I have Tim Apple's home line if you want to call him about something?
<jekstrand> dj-death: Looking at the ANV pipeline libraries MR and I see you're doing an extra indirection for dynamic buffers now.
nchery has quit [Ping timeout: 480 seconds]
<jekstrand> dj-death: That's gonna mean no push constants. :(
<jekstrand> I guess we can optimize to only do the indirection if there's a set missing and tell people to put their dynamic stuff in set0
lemonzest has quit [Quit: WeeChat 3.5]
lynxeye has quit [Quit: Leaving.]
nchery has joined #dri-devel
<jekstrand> Actually, maybe that's ok.... We still generate a per-stage bind_map and UBO pushing requires it go in the bind_map
technopoirot has joined #dri-devel
Haaninjo has quit [Ping timeout: 480 seconds]
<ajax> jenatali: hmph. that gives me build testing but not, like, piglit against xquartz
<ajax> (not to sound ungrateful, thank you for pointing me at it)
<jenatali> ajax: Yeah I have zero context on it other than remembering seeing that MR go by a little while ago :)
<ajax> i keep needing to fix code that has unpleasantly large #ifdef APPLEGL in it and i'd prefer not to break it if i don't have to
<jenatali> ajax: Well, if the build was working, it looks like it at least tries to run meson tests in that pipeline
tjmercier has joined #dri-devel
tjmercier has left #dri-devel [#dri-devel]
tjmercier has joined #dri-devel
shashanks has joined #dri-devel
shashank_s has quit [Ping timeout: 480 seconds]
JohnnyonFlame has joined #dri-devel
sumoon has joined #dri-devel
<dj-death> jekstrand: no really choice I guess (for the indirection)
<dj-death> s/really/real/
Haaninjo has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
<dj-death> jekstrand: and yeah, with independent sets you can't really push much
<dj-death> jekstrand: use push descriptors I guess is the answer
<dj-death> jekstrand: I want to force the push descriptors in the binding table only
<dj-death> for some other work
heat_ has joined #dri-devel
heat has quit [Read error: Connection reset by peer]
<jekstrand> dj-death: So I think push is actually going to be fine the more I think about it.
<jekstrand> dj-death: Because push only happens for stuff that ends up in the binding table and, since that's per-stage, independent sets don't matter. All that matters is that you have all the descriptor set layouts for the given stage.
toolchains has joined #dri-devel
<dj-death> jekstrand: hmmm independent sets are within a stage
<dj-death> jekstrand: oh I see
<dj-death> yeah we do have one bind map per stage
<dj-death> so should be find
<jekstrand> dj-death: I'm re-thinking through all this stuff for NVK.
<jekstrand> Trying to build it with all the flexibility up-front.
<dj-death> I did disable the binding table for independent sets :(
<jekstrand> Why?
<dj-death> but actually I think it only applies to dynamic buffers
<dj-death> because I didn't know better
<jekstrand> Binding table should work fine with independent sets
<jekstrand> Ah
<jekstrand> Yeah, it took me a bit to wrap my head around the independent set stuff too
<dj-death> because the offsets are in the flatten table
<jekstrand> And a chat w/ Tobias to make sure I wasn't crazy. :)
<airlied> jekstrand: would love to push desc sets below galliim for lvp, but uggh
<dj-death> jekstrand: also because dynamic offsets are not tracked per stage
<jekstrand> dj-death: That sould be fine, I think.
<dj-death> jekstrand: yeah, not quite sure what I was thinking
<jekstrand> dj-death: It's ok. This extension is thorny
toolchains has quit [Ping timeout: 480 seconds]
<jekstrand> bbrezillon, jenatali: That'll need a pair of dozen patches before it can go in as-is.
pixelcluster has joined #dri-devel
<jenatali> Ack
toolchains has joined #dri-devel
<jekstrand> Other drivers can choose to use it at their leisure but I want the common ref counting for the ones using vk_cmd_queue.
shashank_sharma has joined #dri-devel
<YuGiOhJCJ> hello, I am using a i7-9700 CPU, mesa 22.1.1 and when I check the xorg log file, I see that xorg tries to load the driver "i965_dri.so" that is not anymore provided by mesa, so I am wondering why xorg does not try the iris driver that is an alternative for i965 https://www.phoronix.com/scan.php?page=news_item&px=Mesa-Classic-Retired
<dj-death> jekstrand: yeah I did change a thing at some point to use a single push buffers for all stages, I guess that's going out of the window
shashanks has quit [Ping timeout: 480 seconds]
<dj-death> I kind of like a single buffer
<jenatali> jekstrand: Got a particular timeline in mind for when you want to land that?
shashanks has joined #dri-devel
apinheiro has joined #dri-devel
shashank_s has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
shashank_sharma has quit [Ping timeout: 480 seconds]
shashanks has quit [Ping timeout: 480 seconds]
technopoirot has quit [Ping timeout: 480 seconds]
mvlad has quit [Remote host closed the connection]
mclasen has quit [Ping timeout: 480 seconds]
neonking_ has quit [Remote host closed the connection]
Duke`` has quit [Ping timeout: 480 seconds]
neonking_ has joined #dri-devel
toolchains has joined #dri-devel
<anholt_> YuGiOhJCJ: perhaps you have an ancient x server? xserver should be asking mesa what driver to use as of 8d4be7f6c4f7c673d7ec1a6bfdef944907a3916e
zf has joined #dri-devel
Haaninjo has quit [Quit: Ex-Chat]
toolchains has quit [Ping timeout: 480 seconds]
pixelcluster has quit [Ping timeout: 480 seconds]
saurabhg has quit [Ping timeout: 480 seconds]
gawin has joined #dri-devel
mbrost_ has joined #dri-devel
mbrost has quit [Read error: Connection reset by peer]
ybogdano has quit [Ping timeout: 480 seconds]
gawin has quit [Ping timeout: 480 seconds]
ybogdano has joined #dri-devel
toolchains has joined #dri-devel
ybogdano is now known as Guest3565
ybogdano has joined #dri-devel
mbrost has joined #dri-devel
mbrost has quit [Remote host closed the connection]
Guest3565 has quit [Ping timeout: 480 seconds]
technopoirot has joined #dri-devel
toolchains has quit [Ping timeout: 480 seconds]
pixelcluster has joined #dri-devel
gawin has joined #dri-devel
karolherbst has quit [Ping timeout: 480 seconds]
karolherbst has joined #dri-devel
ybogdano is now known as Guest3566
ybogdano has joined #dri-devel
danvet has quit [Ping timeout: 480 seconds]
technopoirot has quit [Ping timeout: 480 seconds]
Guest3566 has quit [Ping timeout: 480 seconds]
shashanks has joined #dri-devel
<zmike> I assume someone has reported gitlab being crazy today?
<anholt_> #freedesktop admins know, yes.
<zmike> as expected, they're experts
ahajda has quit [Ping timeout: 480 seconds]
<zmike> 🙏
shashank_s has quit [Ping timeout: 480 seconds]
apinheiro has quit [Quit: Leaving]
sul has quit [Ping timeout: 480 seconds]
sul has joined #dri-devel
ybogdano has quit [Ping timeout: 480 seconds]
rasterman has quit [Quit: Gettin' stinky!]
icecream95 has joined #dri-devel
gnuiyl has quit [Ping timeout: 480 seconds]
ybogdano has joined #dri-devel
<anholt_> looks like we're back?
chaim has quit [Quit: Konversation terminated!]
shashank_sharma has joined #dri-devel
shashanks has quit [Ping timeout: 480 seconds]
ngcortes has joined #dri-devel
pcercuei has quit [Quit: dodo]
pixelcluster has quit []