ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
<DavidHeidelberg[m]>
Any idea why only for radeon r300 I have to add `GBM_ALWAYS_SOFTWARE=true` to get `drm-shim` working? All other drivers pass `eglInitialize()` without any issue on CI.
<DavidHeidelberg[m]>
On "runner" with real hardware (e.g. my Intel laptop), there is no need for `GBM_ALWAYS_SOFTWARE=true`
jewins has quit [Quit: jewins]
<DavidHeidelberg[m]>
my assumption is that there is flawed logic in r300 init which won't allow fallback to software when no "real" HW is present
Surkow|laptop has quit [Remote host closed the connection]
co1umbarius has joined #dri-devel
columbarius has quit [Ping timeout: 480 seconds]
Surkow|laptop has joined #dri-devel
yuq825 has joined #dri-devel
danvet has quit [Ping timeout: 480 seconds]
heat has quit [Ping timeout: 480 seconds]
alyssa has joined #dri-devel
<alyssa>
DavidHeidelberg[m]: maybe a bug in r300 drm-shim?
<abhinav__>
vsyrjala sorry for the late response, i was traveling this week. Regarding https://patchwork.freedesktop.org/series/100393/, drm-misc-next is fine and i can see its already applied. Thanks!
Daanct12 has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
<abhinav__>
javierm yes several generations of MSM hw support solid fill planes.
<abhinav__>
pinchartl this optimization has certainly helped us quite a bit in terms of memory bandwidth consumption.
<abhinav__>
regarding daniel's comment, background CRTC color would fill the entire screen with the color
<abhinav__>
plane solid fill would only fill the required ROI and can thus be used even for market applications
<abhinav__>
emersion pinchartl i cannot comment on other hardware, but msm has supported this feature for many generations in our downstream driver and has been a value add feature. I think other hardware could be supporting this as well using a different name? but that being said, it does seem generic enough to be considered for the DRM fwk . the complexity
<abhinav__>
of the core changes are not really that much.
<abhinav__>
*unable to tag danvet here
Daanct12 has quit [Quit: Leaving]
Company has quit [Quit: Leaving]
Akari has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
AndrewR has joined #dri-devel
<AndrewR>
Hello! Is there anything besides stated dependencies for compiling piglit? It seems to stuck for 6 hours doing some python3.9 stuff .. at 1%
Daanct12 has quit [Ping timeout: 480 seconds]
rmckeever has quit [Quit: Leaving]
justsid has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
Daanct12 has joined #dri-devel
mhenning has quit [Quit: mhenning]
lemonzest has joined #dri-devel
Daanct12 has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
Akari has joined #dri-devel
mszyprow_ has joined #dri-devel
dcz_ has joined #dri-devel
alanc has quit [Remote host closed the connection]
<karolherbst>
(which is like the only blocker atm)
yuq825 has left #dri-devel [#dri-devel]
<karolherbst>
ehh... and something else.. need to write the patch for work group info...
Company has joined #dri-devel
<karolherbst>
now I have to think about how to implement CL_DEVICE_LATEST_CONFORMANCE_VERSION_PASSED
srslypascal has quit [Remote host closed the connection]
<DavidHeidelberg[m]>
karolherbst: Matrix client has funny feature, when you posted 🎉 screen got covered by confetti :D
<karolherbst>
lol
<karolherbst>
kusma: ahh you wanted to look into a mastodon instance, or wasn't it you? I know I've talked with somebody about but I forgot who it was
gouchi has quit [Remote host closed the connection]
<eric_engestrom>
karolherbst: it was me, but I haven't had any time to look into it more
<karolherbst>
ahh
<eric_engestrom>
also, huge congrats on conformance!
<karolherbst>
somebody told me there are apparently hosted instances one can sign up on or something
<karolherbst>
tanks!
<karolherbst>
*thanks
srslypascal has joined #dri-devel
<eric_engestrom>
yeah, I talked with the gamedev.place instance admin (aras_p) and he's using masto.host
<eric_engestrom>
that host is not taking more instances right now, but there are others like that
<eric_engestrom>
that way moderation would be the only thing left for us to do
<karolherbst>
I already hear from some admins that you really have to get your infra right if you want to do it seriously, so I kind of prefer having it hosted, but we might also want to know how much money we are able to throw at it
<eric_engestrom>
and we'd keep it invite-only, at least at first, so it would be only existing FDo members who could get an account, which would limit the moderation need quite a bit
<karolherbst>
yeah.. moderation isn't a problem I think :)
<karolherbst>
user or server invite only?
<eric_engestrom>
well, if anyone can join, moderation is a problem, but if it's only among people who respect the FDo CoC, it should be fine
<eric_engestrom>
I don't understand your last question?
<karolherbst>
though I wouldn't mind it if it always stays user invite only... you can also make it "registration" based, but users are pain if they have to do something. Like if we want to only allow contributors the way of verifying that sucks
<karolherbst>
eric_engestrom: soo.. you can have the instance that only admin/mods can invite or every user on the instance
<karolherbst>
wouldn't mind user invite only forever tho. That scales good enough
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
bgs has joined #dri-devel
<eric_engestrom>
yeah, I agree
<alyssa>
which one lets you delete clover fastest
<alyssa>
er
<alyssa>
whoops
<alyssa>
eric_engestrom: maybe a bit gatekeepy but 'an x.org member'? ..
<eric_engestrom>
alyssa: yeah, that might be a good criteria
<eric_engestrom>
and gatekeeping is necessary at first, the question is how long until we open the gates and how much
karolherbst has quit [Remote host closed the connection]
karolherbst has joined #dri-devel
<alyssa>
I'm not used to it yet
Jeremy_Rand_Talos_ has quit [Remote host closed the connection]
Jeremy_Rand_Talos_ has joined #dri-devel
srslypascal is now known as Guest1298
srslypascal has joined #dri-devel
Leopold_ has quit [Remote host closed the connection]
Guest1298 has quit [Ping timeout: 480 seconds]
Leopold_ has joined #dri-devel
heat has joined #dri-devel
tagr has quit [Remote host closed the connection]
mszyprow_ has joined #dri-devel
gouchi has joined #dri-devel
mhenning has joined #dri-devel
tagr has joined #dri-devel
<karolherbst>
I wonder how long it will take to troll google into using rusticl instead.... lol
Leopold_ has quit [Remote host closed the connection]
<jenatali>
Instead of clspv?
<karolherbst>
yeah
<karolherbst>
or well.. clvk
<kisak>
instead of tricking, be better than the competition?
<karolherbst>
thing is.. I already solved problems clvk still struggles with :P
<karolherbst>
though I suspect at this point clvk is the more stable solution, because it actually runs on quite a few vulkan impls
Leopold_ has joined #dri-devel
bmodem has quit [Ping timeout: 480 seconds]
rasterman has joined #dri-devel
<alyssa>
Does anyone know what happens to out-of-bounds values written to pure integer framebuffers in OpenGL?
<alyssa>
i.e. fragment shader does `highp out int c; c = 260;` with a GL_R8UI format
<alyssa>
clamp? wrap? undefined?
<alyssa>
I can't find anything in the spec about this
<HdkR>
I always thought it would clamp but I have no idea now
<HdkR>
Actualy Vulkan defines it as undefined. Maybe that matches GLSL behaviour :D
sysescool has quit [Remote host closed the connection]
sysescool has joined #dri-devel
mszyprow_ has quit [Ping timeout: 480 seconds]
<jenatali>
D3D says clamp, FWIW
<HdkR>
Looks like the original EXT_gpu_shader4 spec doesn't state anything about it and that was carried through to the final spec to not be defined.
<jenatali>
"For all formats, there is a clamp to the RenderTarget range after blend, before writing values out to memory."
<alyssa>
jenatali: I always thought it was clamp in GL too but I cannot find a citation for that
<airlied>
pretty sure i fixed bugs in drivers before to clamp it
<airlied>
some old radeknsi hw did mask
<jenatali>
In general, unless the specs say different things, you can assume for any desktop GPU that it'll follow the D3D spec. Usually it matches GL behavior except for some corner cases
<alyssa>
Right, Mali clamps in hw too (at least Bifrost onwards)
<jenatali>
But yeah when it comes to mobile I dunno
<alyssa>
But for Asahi, I get to implement the whole blend pipeline in shader code
<alyssa>
so I get to make these choices :-p
<HdkR>
Time to make an ES 3.0 test that overflows the output and checks what the result is? :P
<alyssa>
HdkR: Hmm?
<HdkR>
Since it seems underspecified, might be fun to see if behaviour differs between stacks
<alyssa>
Ah, sure
<alyssa>
airlied: I'm also pretty sure I fixed Midgard to clamp instead of mask but now I'm doubting myself
<airlied>
something in vk cts czred
<airlied>
cared
<airlied>
might be a blit meta path
<airlied>
maybe check blit language
* alyssa
might believe blits being different
<alyssa>
meanwhile clamping is causing some tests to fail, umm