06:41 diagonal2x: so technically it's correct 000
07:19 diagonal2x: +++
11:49 diagonal2x: Actually guys in fact there is a better way to do it, when arriving to minimum there is a way to subtract from half a value, and check the common value which is half of that half, and remove this, and in hash compensate back, then there's enough room cause high order bits are now removed, it's just two more instructions to lift to compressed format, I checked the logics behind my first pass, so it's fully solved now, so the lifter is fully
11:49 diagonal2x: dma compatible.
14:46 hakzsam: agd5f_: hi, looks like AMDGPU broke my pitcairn GPU, here's the dmesg https://pastebin.com/raw/NUrR6zZq fw issue?
14:48 hakzsam: according to linux-firmware, pitcairn fws didn't change
14:50 agd5f: hakzsam, we haven't touched any SI code in ages
14:51 agd5f: maybe some problem with your local FW image?
14:52 hakzsam: I don't think so, the problem happens on two different machines
14:52 hakzsam: and it happened recently, like couple of weeks ago
14:52 hakzsam: maybe after a system update, I don't quite remember
14:57 hakzsam: I'm going to check older kernels
14:59 hakzsam: with a random commit from 2 months ago, I have a differen error https://pastebin.com/raw/sj8gmVE7
15:04 agd5f: that was fixed with this commit: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=219223eca4d8b39486107c62a52c9667016aa3b9
15:05 hakzsam: ack
15:07 agd5f: Only thing I can think of that might be related is: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=037b98a2312e2587163de14afae8ae1b64b67dda
15:09 agd5f: ultimately: 56e449603f0a ("drm/sched: Convert the GPU scheduler to variable number of run-queues")
15:10 agd5f: but none of that should affect hw state
15:10 hakzsam: with 2dde18cd1d8f it doesn't even boot, black screen, no ssh
15:12 hakzsam: trying 930ef5b35200e8c521ed8999c7c1c8ccf4164a0c now
15:23 hakzsam: same as the previous one :(
15:32 hakzsam: agd5f: I checkout'd the commit right before 037b98a2312e2587163de14afae8ae1b64b67dda and I have the same uvd fw issue btw
15:34 hakzsam: but HEAD of amd-staging-drm-next is also broken
15:54 agd5f: hakzsam, try reverting: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=037b98a2312e2587163de14afae8ae1b64b67dda
15:54 agd5f: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=dab96d8b61aab1a4f99d0b86964a6c40e7bb1756
15:54 agd5f: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b70438004a14f4d0f9890b3297cd66248728546c
15:54 agd5f: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=56e449603f0ac580700621a356d35d5716a62ce5
16:07 hakzsam: agd5f: for some reasons, I can boot with 6.6.5-arch1-1 (stock archlinux kernel) but I get https://pastebin.com/raw/1dSxgNUR
16:07 hakzsam: I'm not sure what's going on...
16:07 agd5f: hakzsam, using radeon
16:08 hakzsam: oh right, I didn't add the options there
16:10 hakzsam: okay with options added, black screen, no ssh again
16:12 hakzsam: I tried 6995e2de6891c724bfeb2db33d7b87775f913ad1 (6.4), same issue
16:13 hakzsam: maybe the gpu itself is broken now? first time I have such an issue though
16:20 hakzsam: is there a way to skip initializing this uvd block completely?
16:35 johnny0: hakzsam: you can use the kernel module parameter amdgpu.ip_block_mask=0xffffff7f to disable ip block 7 (which is UVD on your card)
17:22 hakzsam: that worked
17:24 hakzsam: I mean, it booted and launched amdgpu with that option but it hangs right after
17:33 hakzsam: well, I think that gpu is dead, need to buy a new one