00:13 sarthakbhatt1: test
00:19 sarthakbhatt1: test
00:33 sarthakbhatt1: test
00:38 sarthakbhatt1: test
00:38 HdkR: A peculiar way to test if an IRC connection is working
00:43 sarthakbhatt1: I'm sorry for bothering the team members, I was trying to integrate irc with discord using matterbridge. I tested with other channels and its working but this is still only sending me message in discord but cant send it discord->irc
00:44 Sachiel: probably missing nickserv identify
00:45 sarthakbhatt1: I already added my nick to access so that it can bypass the identify process
00:45 daniels: sarthakbhatt1: your messages are visible on IRC
00:46 sarthakbhatt1: yes but sending from discord to irc is not going on.
00:48 sarthakbhatt1: I can read every message from the irc in discord. matterbridge setup is successful with irc
00:53 sarthakbhatt1: okay I should stop anoying others now! sorry
00:54 airlied: we have a discord bridge for nouveau, karolherbst set it up
03:06 jenatali: sarthakbhatt1: "bypass the identify process"? I don't think that's a thing...
03:06 jenatali: FWIW I can see your messages coming through Matrix even if they're not making it to IRC
03:10 sarthakbhatt1: okay then I think its the identify thing that is stoping my request to irc.
03:13 HdkR: +M blocks all messages from unidentified users yes
10:41 karolherbst: jenatali: how long are you on vacation? I want to merge https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27663 (it's the opencl-c.h thing and other stuff)
11:51 saneloops: vsyrjala: sorry for highlighting, common sense would indicate that bus is formed so, so the bit field is split into two or four, so in case of two equal fields max combinations of both fields is 256, but for every digits in 256, so the maximum comes as 256+256+256+256 but wieghted maximum comes as 256*256*256*256 well minus one too there, so with no collisions you add wights to those digits to help translation, but hardware inside it's
11:51 saneloops: block can do whatever it wishes until their topology works, i only speculate that both of my examples are for 32bit bus widths , in other words modulating over two or four fields does not make it 16 or 8 bit bus algorithms. Higher math i took few and succeeded few in, but it's actually amazingly enough quite logical and also imo, linux is just amazing and we have those places to start to do better performance too.
12:12 saneloops: so 64bit friendly non colliding weights would be 64to128 for every 64 powers
12:12 tzimmermann: javierm, have a merged your sparc fix yet?
12:13 tzimmermann: s/a/you/
12:16 saneloops: we can simply get there on any system, but i like to work under unix, i came so smart under those platforms and maintain my go under such, cause there isn't this commercial garbage and virus headaches, and system bothering me, yet so good system, so it's pleasure to work still, and yep, i can make some modern design under linux , but i use some other third party code to get it done quicker.
12:26 javierm: tzimmermann: I have not, but since is for sparc I didn't know if you push through drm-misc-fixes...
12:27 javierm: *could push
12:27 saneloops: so in the end you take one checkpointint external solution or debugging, and debug a little like testing the piano (i do not play piano but i play keyboards), then you modify a bit , and compose a bit, cause some kits are nearly ideal, so all becomes realistic that way, without major headaches, checking the documentation of all, and doing what you love to do.
12:28 tzimmermann: javierm, yes i can put it into drm-misc-fixes.
12:29 tzimmermann: i'm going to add a cc fot stable@ as well
12:33 javierm: tzimmermann: makes sense. Lately I haven't added Cc: stable@ anymore since it seems that anything with Fixes: is picked by them
12:34 javierm: tzimmermann: and thanks! As mentioned I didn't push myself because wasn't sure if had to wait for a sparc maintainer to do it
12:35 tzimmermann: is there still a sparc maintainer? the recent video changes in arch/sparc have been reviewed by arnd
12:37 javierm: tzimmermann: I don't know tbh...
12:40 javierm: tzimmermann: but yeah, I guess nobody will complain if we push it through drm-misc and I will be suprised if there are conflicting changes in a sparc tree (if this exists)
12:46 arnd: tzimmermann: Andreas Larsson has just signed up as the new maintainer for sparc, with his main job being the cobham-gaisler leon (sparc32) support
13:03 stalkerg: friedrich: sorry for the random question, do you know what exactly different between RDNA2 and RDNA3 from RT point of view (is it possible to find docs)? Did RADV support a new instructions or not?
13:08 jenatali: karolherbst: I'll be back Tuesday
13:12 stalkerg: jenatali: thanks
13:12 stalkerg: ops :(
13:32 tzimmermann: javierm, merged
13:32 tzimmermann: arnd, thanks
13:42 jenatali: karolherbst: as long as it doesn't break CI consider it acked. I looked close enough to agree it shouldn't be problematic from a compile time perspective
13:43 karolherbst: cool, thanks :)
13:46 javierm: tzimmermann: thanks
14:34 zmike: mareko: am I correct in reading the spec that there are zero limitations on attrib stride other than it cannot be negative?
14:51 mareko: zmike: yes
14:51 mareko: zmike: and there is queriable maximum
14:51 zmike: right
14:56 zmike: mareko: alright, apparently I was gaslighting myself and rgba8 with stride=1 is currently allowed
14:56 zmike: same with rgba16/2
14:59 mareko: zmike: in Vulkan?
14:59 zmike: yes
15:00 mareko: well, that's at least something
15:01 mareko: hypothetically an extension can be added to Vulkan to allow strides with nonsense alignment, and RADV can handle that
15:01 zmike: hypothetically if such an extension were going to be added I would already be adding it
15:02 mareko: is it not possible?
15:02 zmike: it's possible
15:02 mareko: radv has to lower unaligned stuff because it can't do rgba8/1
15:03 mareko: any new alignment rules is just incremental work
15:24 jenatali: zmike: stride=1 for r16?
15:25 jenatali: We added new alignment rules to D3D but the spec says that the final address for any vertex component needs be aligned
15:25 zmike: stride=1 for r16 is not allowed
15:25 jenatali: Oh I see, the /2 at the end
15:26 jenatali: Reading fail :)
15:26 zmike: vacation fail
15:26 jenatali: Yeah that too
16:50 Hazematman: Hey, is there an easy way to check why a symbol is not being included in my mesa build? For some reason glEGLImageTargetTexture2DOES is not included when I build mesa myself even though glxinfo/eglinfo reports GL_OES_EGL_image_external as a supported extension
16:54 soreau: to what prefix did you install mesa?
16:56 Hazematman: to a custom one, and when I run I set LD_LIBRARY_PATH and LIBGL_DRIVERS_PATH to point to the lib and dri folder in there respectively
16:56 daniels: Hazematman: you need to look it up via eglGetProcAddress, not call it directly
16:56 soreau: or use libepoxy?
16:57 Hazematman: daniels: That's what I'm doing eglGetProcAddress is returning null
16:59 soreau: make sure you're pointing LD_LIBRARY_PATH to the right location too, usually it's $prefix/lib/$libdir/, not just $prefix/lib/
17:00 soreau: and if you set LD_LIBRARY_PATH, IME, LIBGL_DRIVERS_PATH isn't needed
17:00 tnt: That could also depend of the version of the context you requested right ?
17:03 Hazematman: for context on what I'm doing. I'm trying to finish zmike's work here https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16224 and I'm using a slightly modified version of this application to test dmabuf import/export across different drivers https://gitlab.com/blaztinn/dma-buf-texture-sharing
17:04 Hazematman: for some reason with my modified version of mesa I can't get the glEGLImageTargetTexture2DOES symbol, so I'm trying to figure out if I need some build system setting or gallium pipe cap to get it to be available
17:06 daniels: nope, you definitely don't, so something strange must be going on with either your build or how you're setting up your context in your app
17:06 daniels: all Wayland compositors use that entrypoint, so if you can run Weston then it is there
17:07 zmike: for testing I would just run weston in one driver and weston-simple-dmabuf-egl on another
17:07 tnt: Hazematman: "modified" how ?
17:12 Hazematman: tnt: You can look at the MR, as that has the majority of the changes. But my changes on top of it are only to `llvmpipe_resource_from_handle` & `llvmpipe_resource_get_handle` to allocate and import real dmabuf, instead of treating them like opaque fds
17:13 Hazematman: <daniels> "all Wayland compositors use that..." <- something is probably just mess up in my build, I'll do more digging into that. Thanks for confirming!
17:13 daniels: np :)
17:16 soreau: since it's zink, does it possibly need VK_ICD_FILENAMES too?
17:21 Hazematman: soreau: I'm not testing zink atm, just llvmpipe
17:47 abhinav__: vsyrjala jani Hi, any feedback on https://patchwork.freedesktop.org/patch/579121/?series=130145&rev=1 ?
18:45 saneloops: I won't request attention anymore, but it's that my mind has been somewhere in the toilet pot and when i discover something i try to request some help, i feel roller coaster ride type of nerves tbh. and i put those libraries sometimes to request help in achieving something collectively cause i can not pull myself enough together, in other words my nerves fall so often i think i am not competent enough. But tbh. if someone else can not
18:45 saneloops: also understand like me, it is fine, but if you tried it's better than fine, and if you made something special its super fine. In other words me as nerve wreck my own don't judge critically.
18:46 saneloops: i am still working on smallbitset reading
18:46 saneloops: and discover something over time and time
19:17 saneloops: technically should also log onto rust, but if anyone finds out who i am under any channel, i get banned at instant. So its better and makes more sense to read the docs, more effective.
19:27 saneloops: imo there is still no such language way of understanding the code, since first thing you notice on the quite rich set of tests, there is no add(...) set that is not in growing order, and every set of bitsets goes to its own variable if needed to have repeated values
19:27 saneloops: only the subset instance is capable of adding many sets through vec mapping
19:29 saneloops: what i am trying to say , is that sometimes at the end of the day, its numbers that count
19:36 saneloops: hence the first really dilemma is how those two iterators in tandem really work, and what happens if after those iterations you release the mapping vector
19:36 saneloops: but this solution appears as where compilation and runtime is merged together, i took them as separate in my vision
19:39 saneloops: overall goal is to understand how rust marks the static and such variables, i got into position of const vs varying but that was clear, there is mutability keyword
19:41 saneloops: but i already arrived into the paragraph in that book, they have something to mess with the liveness
19:42 saneloops: and order of those live variables another keyword