ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
<mareko> alyssa: clamp
<alyssa> mareko: any idea where the spec says that?
<mareko> no, but our hw clamps
<alyssa> ours too, I know clamping is allowed, I just don't know if it's required
<mareko> uint format - same as clamping with umin/umax, int format - same as clamping with imin/imax
pochu_ has joined #dri-devel
pochu has quit [Ping timeout: 480 seconds]
<alyssa> If I *don't* insert clamping code dEQP-GLES3.functional.draw_buffers_indexed.random.max_required_draw_buffers.1 is passing, but it fails if I do, wtf
<alyssa> incidentally that's failing on both mali-t860 and softpipe
<mareko> that test is skipped on radeonsi
pochu has joined #dri-devel
<mareko> ./modules/gles3/deqp-gles3 -n dEQP-GLES3.functional.draw_buffers_indexed.random.max_required_draw_buffers.1 ... print 0/0
<mareko> *prints
<alyssa> it used to be dEQP-GLES31.functional.draw_buffers_indexed.random.max_required_draw_buffers.1
<alyssa> they moved it to GLES3 because it doesn't actually depend on any 31 functionality
<mareko> ok, that passes
pochu_ has quit [Ping timeout: 480 seconds]
<mareko> the only thing that can be wrong with clamping is that you clamp as signed int when the format is unsigned
<alyssa> mmhh
<alyssa> I see what I got wrong
<alyssa> ld_tile zero extends instead of sign extends, I think
<alyssa> on agx
<alyssa> grumble
<HdkR> yay for additional sign extend on load op :D
<alyssa> HdkR: I think it's just extra ALU *sob*
<mareko> also when you convert to 16 bits in the shader and then pass 16 bits to the hw for conversion to 8 bits, the signedness of the 32->16 bit conversion also matters if you clamp there
<alyssa> yep
<alyssa> still easier than Mali ;-p
<alyssa> there we go, passing now
<alyssa> with clamp
<alyssa> good bug fix but dEQP-GLES3.functional.fragment_out.random.51 is still stubbornly failing :v
pcercuei has quit [Quit: dodo]
Jeremy_Rand_Talos__ has quit [Remote host closed the connection]
Jeremy_Rand_Talos__ has joined #dri-devel
columbarius has joined #dri-devel
co1umbarius has quit [Ping timeout: 480 seconds]
apinheiro has quit [Quit: Leaving]
mhenning has quit [Quit: mhenning]
bluepenquin has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
Daaanct12 has quit [Read error: Connection reset by peer]
Danct12 has quit [Read error: Connection reset by peer]
heat_ has quit [Remote host closed the connection]
heat has joined #dri-devel
Danct12 has joined #dri-devel
Lyude has quit [Quit: Bouncer restarting]
Lyude has joined #dri-devel
Leopold_ has quit [Remote host closed the connection]
heat has quit [Remote host closed the connection]
heat has joined #dri-devel
dafna33[m] has quit []
heat has quit [Ping timeout: 480 seconds]
Akari has quit [Ping timeout: 480 seconds]
TMM has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
TMM has joined #dri-devel
kem has quit [Ping timeout: 480 seconds]
kem has joined #dri-devel
pekkari has joined #dri-devel
kem has quit [Ping timeout: 480 seconds]
alanc has quit [Remote host closed the connection]
Duke`` has joined #dri-devel
alanc has joined #dri-devel
dcz_ has joined #dri-devel
kts has joined #dri-devel
kem has joined #dri-devel
lemonzest has joined #dri-devel
pekkari has quit [Quit: Konversation terminated!]
kts has quit [Quit: Leaving]
jolan has quit [Quit: leaving]
jolan has joined #dri-devel
junaid has joined #dri-devel
junaid has quit [Quit: leaving]
junaid has joined #dri-devel
kts has joined #dri-devel
jolan has quit [Remote host closed the connection]
karolherbst_ has joined #dri-devel
karolherbst is now known as Guest1353
karolherbst_ is now known as karolherbst
KunalAgarwal[m][m] is now known as KunalAgarwal[m]
<KunalAgarwal[m]> I am using egl and gles for buffer management and textures. To create a triangle using Legacy GL, do i need indicate somewhere/somehow that I am using legacy GL
Guest1353 has quit [Ping timeout: 480 seconds]
mszyprow_ has joined #dri-devel
tagr has joined #dri-devel
mszyprow_ has quit [Ping timeout: 480 seconds]
<linkmauve> KunalAgarwal[m], you can’t mix and match calls to both APIs, you decide which one you use with eglBindApi() as part of the context creation.
junaid has quit [Quit: leaving]
junaid has joined #dri-devel
<KunalAgarwal[m]> hmm ok. I wanted to test the output side of my code, (without the shaders) so I thought of creating a triangle using legacy GL
kts has quit [Quit: Leaving]
rasterman has joined #dri-devel
Akari has joined #dri-devel
junaid has quit [Quit: leaving]
junaid has joined #dri-devel
kts has joined #dri-devel
junaid has quit [Quit: leaving]
<linkmauve> GLES 2.0+ doesn’t support not using shaders, better write a very simple one (for instance displaying a bright colour, at the given input positions) to make sure that works.
junaid has joined #dri-devel
danvet has quit [Read error: Connection reset by peer]
gouchi has joined #dri-devel
kov has quit [Quit: Coyote finally caught me]
Duke`` has quit [Ping timeout: 480 seconds]
JohnnyonFlame has quit [Ping timeout: 480 seconds]
Duke`` has joined #dri-devel
rasterman has quit [Quit: Gettin' stinky!]
kem has quit [Ping timeout: 480 seconds]
kem has joined #dri-devel
kov has joined #dri-devel
kts has quit [Quit: Leaving]
kts has joined #dri-devel
junaid has quit [Ping timeout: 480 seconds]
bgs has joined #dri-devel
junaid has joined #dri-devel
pcercuei has joined #dri-devel
junaid has quit [Ping timeout: 480 seconds]
danvet has joined #dri-devel
junaid has joined #dri-devel
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
Leopold_ has joined #dri-devel
Akari has quit [Quit: segmentation fault (core dumped)]
Leopold_ has quit []
sysescool_ has joined #dri-devel
Leopold_ has joined #dri-devel
sysescool has quit [Ping timeout: 480 seconds]
kts has quit [Quit: Leaving]
bmodem has joined #dri-devel
kts has joined #dri-devel
Leopold_ has quit [Remote host closed the connection]
junaid has quit [Remote host closed the connection]
Leopold_ has joined #dri-devel
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #dri-devel
gouchi has quit [Remote host closed the connection]
f11f12 has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
Duke`` has joined #dri-devel
Leopold_ has quit [Remote host closed the connection]
mszyprow_ has joined #dri-devel
Leopold_ has joined #dri-devel
ElementW has quit [Ping timeout: 480 seconds]
zehortigoza has quit [Remote host closed the connection]
<alyssa> KunalAgarwal[m]: linkmauve is correct, and more to the point, mesa translates fixed function GL(ES) into shaders internally
<karolherbst> and " 8 files changed, 986 insertions(+), 69 deletions(-)" later SVM was implemented
<alyssa> karolherbst: Oof!
<karolherbst> most of it is just API boilerplate code tho
<karolherbst> and dealing with extension stuff and all of that
<alyssa> karolherbst: + // SAFETY: it's undefined behavior if the application screws up
<alyssa> *upside down face*
<karolherbst> :D
<karolherbst> I have better ones
<karolherbst> "// SAFETY: Just hoping the application is alright." 🙃
<karolherbst> but yeah.. I just handwave on everything the OpenCL spec demands
<alyssa> Yep
<karolherbst> I am wondering if I get away by requiring system SVM from all the drivers
<karolherbst> because that's what developers want to have anyway
<karolherbst> the alternative would be to rewrite all of the existing drivers not supporting that
<jenatali> FWIW if you ever want rusticl on Windows you'll need to deal with it not being there
<karolherbst> yeah well.. SVM isn't advertised then
<karolherbst> not much I can do about it then
<jenatali> Oh I see, you mean "system SVM" as a type of SVM, not just as having it at all
<jenatali> Nevermind :) thought you were saying rusticl was going to require SVM
<karolherbst> ahh no
<karolherbst> I meant only bothering abuot fine grained system SVM
<karolherbst> because the other levels are really painful to deal with
<jenatali> Looks like Nouveau is the only driver that reports SVM at all?
<karolherbst> yeah
<karolherbst> for.... "reasons"
<karolherbst> you remember Nvidia pushing for HMM to be included into the linux kernel? :D
<karolherbst> though I really have to check out what needs to be done inside iris and radeonsi to enable it there as well
<jenatali> No I haven't paid attention to that :P
<karolherbst> :D
<karolherbst> but you are aware of that rule that kernel UAPIs or well.. whatever kind of require open source userspace?
<karolherbst> go figure :P
<jenatali> Yep
JohnnyonFlame has joined #dri-devel
MayeulC[m] has joined #dri-devel
<jenatali> karolherbst: You might care about https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19709 if you're using the scale_fdiv pass (which I think you are?)
Leopold_ has quit []
Haaninjo has joined #dri-devel
srslypascal is now known as Guest1391
srslypascal has joined #dri-devel
mszyprow_ has quit [Read error: Connection reset by peer]
Guest1391 has quit [Ping timeout: 480 seconds]
Duke`` has quit []
JohnnyonFlame has quit [Ping timeout: 480 seconds]
ElementW has joined #dri-devel
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
<jenatali> MayeulC: Your messages aren't making it to IRC because you're not authorized with NickServ
rgallaispou has joined #dri-devel
lemonzest has quit [Quit: WeeChat 3.6]
kts has quit [Quit: Leaving]
rgallaispou has quit [Read error: Connection reset by peer]
JohnnyonFlame has joined #dri-devel
co1umbarius has joined #dri-devel
Leopold_ has joined #dri-devel
<alyssa> jenatali: we're using it, yeah
<karolherbst> jenatali: I'd regression test that MR, but somebody broke iris 🙃
columbarius has quit [Ping timeout: 480 seconds]
<jenatali> :(
<karolherbst> though it regressed a few days ago, so I can probably rebase
<alyssa> jenatali: r-b, I keep forgetting there's a good reason to not just .src.ssa
<jenatali> Yeah, same (apparently)
<jenatali> This isn't the first time I've gotten vector/swizzle lowering wrong and it won't be the last lol
<alyssa> Yeah...
<alyssa> For stuff that only works on scalars, nir_ssa_scalar is the way to go
<alyssa> alyssa@maud:~/mesa/src$ git grep 'src.ssa' | wc -l
<alyssa> 595
<karolherbst> "cool"
<alyssa> jenatali: I don't think I want to bet on scale_fdiv being the only bug here ;p
<jenatali> Same
<jenatali> I wonder if this is the reason Geekbench CL tests are getting wrong answers... I should try that soon
<jenatali> alyssa: You probably want to grep for src[index].src.ssa. Using intrinsic srcs directly is probably fine
MayeulC[m] has quit []
MayeulC[m] has joined #dri-devel
MayeulC[m] has quit []
MayeulC[m] has joined #dri-devel
MayeulC[m] is now known as MayeulC
<alyssa> jenatali: Mm, right
Akari has joined #dri-devel
MayeulC has quit []
MayeulC has joined #dri-devel
<alyssa> jenatali: opt_if looks suss
<jenatali> Maybe make an issue with a checklist of passes to verify?
* alyssa shrugs
<alyssa> That takes time and energy that i do not in abundance right now, heh
<jenatali> Valid
MayeulC is now known as Guest1399
MayeulC has joined #dri-devel
MayeulC has quit []
MayeulC has joined #dri-devel
MayeulC has left #dri-devel [#dri-devel]
Guest1399 has quit []
enick_357 has joined #dri-devel
enick_357 has quit []
enick_357 has joined #dri-devel
enick_357 has quit []
enick_357 has joined #dri-devel
enick_357 has quit []
enick_357 has joined #dri-devel
MayeulC has joined #dri-devel
<MayeulC> That verification thing is new to me
<MayeulC> Anyway, looks like I left the Matrix bridge in a weird state for my account, maybe rate-limitted by NickServ. We'll see later.
<MayeulC> This is the closest to a mesa or radv channel, right? per https://docs.mesa3d.org/lists.html
<jenatali> Yep
<pixelcluster> There's also #radeon for radv/amd stuff
<MayeulC> Ah, thanks. Well, I guess this is a bit of a generic question still: I'm trying to have the mesa git driver expose VK_NV_device_generated_commands, but it seems it will not for some reason (halo infinite refuses to work). Any way to force that with an env var? It seems MESA_EXTENSION_OVERRIDE may not be applicable to Vulkan extensions?
<MayeulC> (that version was just compiled with the Arch PKGBUILD, this extension is only enabled for vk3d with driconf: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17269/diffs#d301a3059c2e4c91694d5eb7f1f0558d6a11b4fa_42_42 )
<MayeulC> I've been looking into driconf, but it's unclear if it can be overriden via a local config file.
<pixelcluster> RADV doesn't fully support the extension, it only supports the subset relevant to vkd3d-proton's implementation of ExecuteIndirect, that's why it is enabled for vkd3d
<pixelcluster> but since Halo Infinite runs through vkd3d, that shouldn't be a problem, are you sure that extension support is causing the issue?
<MayeulC> I am not sure, but the symptoms are the same (game error message "missing feature: Hardware tiled Resources Tier 2")
<bnieuwenhuizen> that is sparse textures, not ExecuteIndirect
<bnieuwenhuizen> which GPU is this
<MayeulC> Ah, that is good to know, thank you.
<MayeulC> That's a R9 Fury. I was able to run Halo Infinite with the mesa-git flatpak extension before the latest update
<bnieuwenhuizen> I think we disabled sparse texturing on that GPU because we still saw hangs with CTS in sparse texture tests
Lyude has quit [Read error: Connection reset by peer]
<MayeulC> However the game now requires a valve-provided proton "Hotfix" branch. For some reason valve-provided proton will not work well with the mesa-git flatpak extension: https://github.com/flathub/com.valvesoftware.Steam/issues/1006
<MayeulC> bnieuwenhuizen: thanks for the heads-up. I only tried basic gameplay with that game, but never encountered a hang. Not sure if I have access to the CTS.
<bnieuwenhuizen> You can change https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/amd/vulkan/radv_device.c#L1324 to something earlier to enable sparse textures
<bnieuwenhuizen> interesting because even on newer gens there is a known hang issue :)
<MayeulC> I do get complete, unrecoverable system lockups on some games though ( https://gitlab.freedesktop.org/mesa/mesa/-/issues/864 ), but I chalked it up to some hardware defect: this is a second-hand GPU after all
<MayeulC> bnieuwenhuizen: Would you be interested in my further testing this feature on my GPU? Is there a reproducer for that hang? I'll try to change that line, thanks a lot for the pointer.
<bnieuwenhuizen> MayeulC: we have someone with the right GPU, just haven't found the time
heat has joined #dri-devel
bmodem has quit [Ping timeout: 480 seconds]
Lyude has joined #dri-devel
gouchi has joined #dri-devel
<MayeulC> I'm going to try that, thanks... I'm still idling around with a Matrix client, feel free to ping me.
MayeulC has left #dri-devel [Konversation terminated!]
Leopold_ has quit [Remote host closed the connection]
sravn has quit []
mhenning has joined #dri-devel
Leopold_ has joined #dri-devel
f11f12 has quit [Quit: Leaving]
TMM has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
TMM has joined #dri-devel
<karolherbst> jenatali: host_ptr are not supposed to be coherent or anything, right?
<jenatali> Nope
<karolherbst> okay...
<jenatali> Needs to be explicitly mapped and unmapped
<karolherbst> that's good, just means that nouveau us broken 🙃
<karolherbst> apparently nouveau doesn't return the host ptr when mapping
<karolherbst> which is normally fine, except for SVM where the CTS is doing silly things anyway
<karolherbst> but I think the CTS requires the ptr to match up anyway
<karolherbst> I think I'm mainly confused why this all worked with clover before.. oh well
<karolherbst> ohh no.. my path for GPUs with discrete memory is just screwed up, nice... another bug found
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #dri-devel
sravn has joined #dri-devel
apinheiro has joined #dri-devel
<LaserEyess> some drivers don't support the DRM atomic API, and new drivers are supposed to support it
<LaserEyess> but will never drivers support the legacy API too? As in, do they get it For Free™ just by supporting atomic?
<emersion> LaserEyess: yes, they get it for free
<emersion> (mostly)
<emersion> there are a few quirks, but core DRM converts legacy to atomic
<LaserEyess> sweet
<LaserEyess> are these quirks things that would effect, for example, displaying a dumb buffer to the screen?
dcz_ has quit [Ping timeout: 480 seconds]
<emersion> nah, they're just things that the legacy uAPI does differently
<emersion> like cursor updates, and async page-flips
<karolherbst> "fun" libasan + SVM doesn't work so well together
<LaserEyess> emersion: awesome, thanks
fab has quit [Quit: fab]
gouchi has quit [Remote host closed the connection]
Duke`` has joined #dri-devel
bgs has quit [Remote host closed the connection]
Duke`` has quit [Ping timeout: 480 seconds]
Jeremy_Rand_Talos__ has quit [Remote host closed the connection]
Jeremy_Rand_Talos__ has joined #dri-devel
Haaninjo has quit [Quit: Ex-Chat]
pcercuei has quit [Quit: dodo]
srslypascal is now known as Guest1415
srslypascal has joined #dri-devel
Guest1415 has quit [Ping timeout: 480 seconds]