11:19 haasn: I accidentally ran my test framework against nouveau and stumbled across an issue where (reproducibly) 3D rgba16 texture downloads to buffers did not work (the buffer still contained all zeroes after the fact)
11:19 haasn: I have no idea why the issue was this oddly specific, 3D downloads of other texture formats worked fine, and 3D download of textures to pointers (not buffers) also worked fine
11:32 karolherbst: haasn: on what GPU?
11:32 karolherbst: but yeah.. the way nouveau deals with 3d is highly "weird"
11:33 karolherbst: most of the programming is done in 2d space and you have to deal with 3d by either offseting the base pointer or specifying the z coord, which is often not possible
11:33 haasn: 01:00.0 VGA compatible controller: NVIDIA Corporation GF116 [GeForce GTS 450 Rev. 2] (rev a1)
11:33 haasn: It may be worth pointing out that even the z=0 slice was all zeroes
11:34 karolherbst: okay, that's a bit odd
11:35 karolherbst: haasn: antyhing in dmesg?
11:36 haasn: I can check later
11:36 haasn: Kinda annoying to set up the machine again
11:37 karolherbst: might also be a stupid flushing issue...
18:07 Wally: -
19:05 Wally: karolherbst: Was the v3 of your PostRADualIssue pass complete?
19:12 karolherbst: maybe, but I think it was causing some bugs _somewhere_
19:12 karolherbst: maybe I fixed that though
19:19 Wally: Not in v3 im guessing?
19:51 karolherbst: mhh, not sure
19:51 karolherbst: my latest branch is probably this: https://gitlab.freedesktop.org/karolherbst/mesa/-/commits/dual_issue_v3/
19:51 karolherbst: I think it picked _some_ patches