05:57 zzoon[m]: Lynne: have a question about pMiCol/RowStarts of StdVideoAV1TileInfo .
05:57 zzoon[m]: the unit of them should be MI not SB IIUC.
05:57 zzoon[m]: but current implementation seems to use them as a SB unit.
05:58 zzoon[m]: what do y ou think? am i misunderstanding something?
06:43 zzoon[m]: I think it should be https://pastebin.com/E2JZ64YD
08:06 Lynne: zzoon[m]: I thought it was already in SB units?
08:06 Lynne: oh, you're right
08:38 tzimmermann: lumag, ping
08:53 lumag: tzimmermann: r-b and applied
08:54 tzimmermann: lumag, nice. thanks.
08:57 zzoon[m]: Lynne: yeah.. radv and anv also consider sb unit for them.. which should be fixed too.
08:58 Lynne: thanks, could you send a patch?
08:58 zzoon[m]: yeah sure.
08:58 Lynne: also could you look into the h264/hevc/av1 seeking corruption on intel?
08:58 Lynne: it's real bad, I have to use vaapi
09:00 zzoon[m]: I'm on https://gitlab.freedesktop.org/mesa/mesa/-/issues/14634 now.
09:00 zzoon[m]: if there are other corruption issues.. please file it and provide test media.
09:05 zzoon[m]: or you can tell me directly of course.
09:46 Lynne: zzoon[m]: https://files.lynne.ee/intel_corruption.ts
09:46 Lynne: it can be replicated on all codecs
09:46 Lynne: but its most visible on h264
09:46 Lynne: it started back in october
09:48 Lynne: you can replicate it by playing the file in mpv
09:49 zzoon[m]: ok thanks
18:41 dcbaker: hakzsam: I've got 47366527ce20b54d1b1a4811bcb6481e09a38402, nominated for 25.3.x. It applies cleanly, but there's 3 undefined functions that don't seem to be simple renames, and I'm not sure what should be pulled in or not there. I've also got 76e7a1cde4, which has a pretty easy diff, but it creates the entire ac_cmdbuf_sdma.c file, so I wanted to make sure that's what you wanted
18:42 hakzsam: dcbaker: I will check tomorrow, if I forget, please ping again
18:42 dcbaker: mareko: I've got 68690eaf4a which is autonominated as a revert, but it doesn't seem it's really meant for 25.3, I'm going to demoninate it, but if you actually want it let me know
18:42 dcbaker: hakzsam: will do, thanks!
18:50 dcbaker: hakzsam: Ignore that first one, you already fixed that and I didn't do a rebase on the upstream staging/25.3 before I started, so that's my bad
18:51 dcbaker: mareko: I also had to massage "nir/clip_cull_distance_utils: fix assertion failures with GL_EXT_mesh_shader" a bit on the 25.3 branch, I would appreciate if you'd take a look at it and make sure it looks okay
19:04 dcbaker: konstantin: I'm having some issues with "radv: Fix crash if proceed comes before initialize" (58a35647e118f1967f17c5803baa0657da98964c) on 25.3.x, it applies cleanly, but has some missing deps. If I pull in "radv: change the default value of RADV_TRACE_CACHE_COUNTERS on < GFX10" (e03461f3bd432daf6e0bfa31a2e83c3ba7bd8ab1) I can get closer, but I'm still missing some prototypes
19:08 dcbaker: hakzsam: Nevermind on the nevermind, the patch I had in tree doesn't compile. So please look at both of those
19:09 dcbaker: konstantin: The issue I saw with your patch was actually an issue from that bad commit of Samuel's patch, sorry for all of the noise both of you
19:21 dcbaker: hakzsam: One more for you: feed5fbc71
19:28 mareko: dcbaker: yes denominate 68690eaf4a
19:29 dcbaker: mareko: done, thanks
19:31 mareko: dcbaker: "nir/clip_cull_distance_utils: fix assertion failures with GL_EXT_mesh_shader" backport looks good
19:31 dcbaker: thanks!
19:59 sghuge: konstantin: any reason RADV did not land opacity micromap yet?
20:02 sghuge: shadeslayer: planning to work on it for anv.
21:37 ngcortes: qq; what's the easiest way to query the ifwi version of a discrete intel gpu?
21:39 mareko: nir_lower_tex_to_image is on the way
21:40 mareko: hopefully radv and radeonsi will stop supporting a subset of tex opcodes because they will get lowered to image opcodes early
21:55 alyssa: mareko: funny, other parts of the tree lower the other direction