08:53 kode54: whee
08:53 kode54: it didn't die after all
08:53 kode54: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12487
08:54 kode54: and it didn't like being mounted horizontally in such cramped space
09:52 Ristovski: kode54: quite the rabbit hole, that hang with the compute vs gfx clears change is interesting
09:55 Ristovski: What I can't get is why alexs patch targeted gfx9 when mareks patch was for gfx6 (which is what the bug report was for iirc)
09:56 Ristovski: unless those code paths are shared in amdgpu
13:28 pepp: both patches are for gfx9
13:42 Ristovski: pepp: Oh is Raven/Raven2 GFX9 as well?
13:42 Ristovski: it indeed appears to be.. nvm then
13:49 Ristovski: well.. CHIP_RENOIR is a thing.. mareks patch wouldn't match that while the kernel patch would I assume
14:21 Venemo: Ristovski: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/amd/common/amd_family.h#L96 <- these are all the supported GFX9 chips
14:28 Ristovski: Venemo: Indeed, my point was that (unless I am mistaken) '0001-drm-am dgpu-gfx9-manually-control-gfxoff-for-CS.patch' applies the workaround for GFX <9.4.0 (so, vega10 to renoir, inclusive), whereas mesa/3b78dcec only applies to Raven/Raven2 specifically
14:29 Venemo: what is 0001-drm-am dgpu-gfx9-manually-control-gfxoff-for-CS.patch?
14:30 Ristovski: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12310#note_2755501
14:31 Ristovski: though I assume it was just a quick testing patch anyway :P
14:32 Venemo: possibly
15:57 Ristovski: agd5f: is the underlying issue in hw or firmware?
16:40 soreau: I don't know what happened here but after both upgrading mesa and resetting the card manually a handful of times, I had some pretty crazy buffer corruptions. Usually a black cursor on firefox but sometimes the black square would have pixels from another surface from another client (using wayfire)
16:40 soreau: Sometimes coming back from cube screensaver would cause a black glitch on one or more surfaces for a couple frames
16:41 soreau: this lasted for about a week and now (~10 hours ago) I've updated to latest mesa again and things seem fine for the moment
16:41 soreau: but the problem usually doesn't manifest itself immediately, so we'll see
16:42 soreau: buntu kernel 6.8.0-51-generic
21:30 kode54: Ristovski: my issue seems to be with GFXOFF state switching
21:30 kode54: as using the amdgpu.ppfeaturemask to disable it seems to work around the issue