01:32 co1umbarius: is there an example how VK_KHR_sampler_ycbcr_conversion can be used with blit or another copy command?
03:25 Lynne: it can't
03:25 Lynne: you can only use it in a shader
03:25 Lynne: in fact, even using copy or blit commands on multiplane images is sketchy according to the spec
08:17 linkmauve: Am I remembering correctly that not passing DRM_MODE_FB_MODIFIERS to ADDFB2 is guaranteed to be the same as passing in DRM_FORMAT_INVALID with the flag set?
08:18 linkmauve: Or are drivers allowed to treat it differently?
08:28 co1umbarius: Lynne: so the only thing I can do is create a convertion to the target formst
08:29 co1umbarius: at and use an identity shader?
10:32 emersion: linkmauve: the kernel should reject INVALID when the flag is set
10:32 emersion: IOW, the kernel API does not work like the Wayland protocol, INVALID does not mean "implicit modifier" there
10:46 linkmauve: Oh, interesting, thanks!
10:47 linkmauve: Does it make sense to allow INVALID in a libdrm-like library then?
10:54 emersion: i think you need to pick one style or the other
10:55 emersion: I like the INVALID style because it allows to express the legacy path with the same params as the new path
10:55 emersion: but it's purely an aesthetic choice
10:55 emersion: in wlroots, I use INVALID style and then set/unset the flag when crossing to KMS
10:56 linkmauve: In drm-rs we can use Option<Modifier>, so None could mean no modifier, and Some(INVALID) could also be a thing.
10:57 linkmauve: I was wondering whether to accept INVALID there or to outright make it unrepresentable in the type.
10:58 emersion: i think that's a misuse
10:58 emersion: INVALID means "none", or at least "no valid value"
11:00 linkmauve: So None and Some(INVALID) would actually mean the same thing?
11:06 vsyrjala: kms has NONE but that actually means LINEAR. and the NONE name is deprecated
11:08 linkmauve: Of course I meant None as the absence of modifier, not DRM_FORMAT_MOD_NONE.
11:09 vsyrjala: i'd probably not pick such a similar name for it. could be confusing. unless that's forced on you by the language or something
11:13 linkmauve: It isn’t imposed, but that’s an extremely common way in every API in that language to express the absence of a value.
11:14 linkmauve: So people will already be used to None being completely different from Some(DRM_FORMAT_MOD_NONE).
11:40 emersion: i would recommend against exposing the NONE value from the header
11:41 emersion: LINEAR is a much better name
12:51 linkmauve: [184781.690548] sun4i-drm display-engine: [drm:drm_internal_framebuffer_create] bad fb modifier 648518346341351425 for plane 0
12:51 linkmauve: Yet drm_info says it is perfectly ok.
12:52 linkmauve: I’m trying to import a NV12 dmabuf exported from V4L2.
12:52 linkmauve: Using two planes, NV12 format, ALLWINNER_TILED modifier.
12:54 dottedmag: Nicely formatted modifier value btw
12:55 linkmauve: I could fix that in sun4i-drm indeed. :)
12:55 linkmauve: Hex would be much better here.
12:55 emersion: it's in core DRM i believe?
12:55 emersion: (drm_internal_framebuffer_create is the function name)
13:04 linkmauve: Ah, I was passing some wrong modifiers, now I got a fb, yay!
13:58 alyssa: y/quit
21:32 cheako: How do I use drm-xe-next? with 6.5.2-x64v3-xanmod1, do I need any args?
21:33 cheako: I'm trying to follow some instructions from karolherbs, must be the weekend?
21:36 Sachiel: if you have the module built, i915.force_probe=!* xe.force_probe=* in the kernel cmdline
21:39 cheako: Is there a DKMS?
21:41 Sachiel: https://gitlab.freedesktop.org/drm/xe/kernel
21:58 cheako: Thanks
22:38 cheako: It has been 0 days :)
23:40 cheako: No good, with that kernel keyboard and a bunch of other things don't work.