00:43Venemo: Mangix: what are you doing exactly?
01:48Mangix: booting
06:44shelter: i think the latest firmware finally fixed the brightness flicker. it says the fix is for "lower resolutions" but im not sure if 1080p counts as a lower resolution
06:48shelter: https://gitlab.com/kernel-firmware/linux-firmware/-/merge_requests/795
16:40Venemo: Mangix: I mean how did you try to add the firmware?
22:28Mangix: Venemo: copy it to /usr/lib/firmware/amdgpu and also zstd compress it
22:29Mangix: getting amdgpu: amdgpu_vce: Firmware "amdgpu/vce_1_0_0.bin" not found or failed to validate (-19)
22:36Venemo: Mangix: did you regenerate your initramfs after doing so?
22:36Mangix: nope
22:38Venemo: that could be the problem then
22:39Mangix: oh
22:39Mangix: reinstalling the kernel worked
22:39Mangix: [ 9.714456] amdgpu 0000:0d:00.0: amdgpu: VCE busy: VCE_STATUS=0x100, SRBM_STATUS2=0x0
22:40Mangix: vce evclk: 0 ecclk: 0 <--- hmmm
22:42Venemo: reinstalling typically also regenerates the initramfs
22:42Venemo: not sure what those messages mean without context
22:43Mangix: first one was dmesg
22:43Mangix: second was amdgpu_pm_info
22:43Venemo: neither of those look wrong to me
22:43Mangix: they're supposed to be 0?
22:43Venemo: they are clocked when you use the vce, and not otherwise
22:44Mangix: oh OK
22:44Mangix: I assumed it was when the firmware was loaded.
22:45Venemo: yes, it is enabled when it's initialized, then it's clocked down until you use it
22:46Venemo: which gpu do you have exactly, Mangix?
22:48Mangix: 7750
22:49Mangix: https://www.techpowerup.com/gpu-specs/sapphire-hd-7750-oc.b652
22:50Venemo: I've tested on a similar one though not the exact same model
22:51Mangix: stock bios was 800 mhz. I updated it to that card's bios
22:51Mangix: I also added UEFI GOP support to it
22:51Venemo: interesting
22:51Venemo:is not familiar with the tools to do that
22:52Mangix: they're windows tools
22:52Mangix: forgot which GPUs were supported.
22:57Mangix: at least GCN 1.0. I think the main limitation was flash chip size on older GPUs
22:59KitsuWhooa: I have no idea what is being tested/discussed, but seeing that model number I realised I have an `ASUSTeK Computer Inc. Radeon HD 7790 DirectCU II OC` in a VM I could provide remote access to if there's a need
23:01Venemo: 7790 is Bonaire (a 2nd gen GCN chip), not applicable to the current discussion
23:01KitsuWhooa: awh
23:01KitsuWhooa: I used to know the differences back in the day :p
23:02Venemo: the VCE in the 7790 should be already working well and has been for many years
23:02KitsuWhooa: that's nice to hear
23:03Venemo: what we are talking about here is the support for VCE1 (included in some 1st gen GCN chips), which was only recently added
23:03KitsuWhooa: ah!
23:05KitsuWhooa: Yeah, the immediately older card I have is an Evergreen, which definitely won't do
23:06Venemo: I already tested VCE1 on Tahiti and Cape Verde, and Alexandre tested it on Pitcairn, so that covers all chips that have it
23:06KitsuWhooa: oh, cool
23:07Venemo: that said, additional testing is always appreciated, like Mangix is doing here
23:10Mangix: huh. so Bonaire supports VCN then?
23:10Mangix: ah nvm
23:12Venemo: Bonaire has VCE2
23:13Venemo: see https://gitlab.freedesktop.org/agd5f/linux/-/blob/amd-staging-drm-next/Documentation/gpu/amdgpu/dgpu-asic-info-table.csv
23:14Mangix: huh. so what is UVD?
23:14Mangix: is that for decoding?
23:15Venemo: https://en.wikipedia.org/wiki/Unified_Video_Decoder
23:16Mangix: "In early versions of UVD, video post-processing is passed to the pixel shaders and OpenCL kernels." < -- I wonder what this means.
23:17Venemo: on old chips, UVD (unified video decoder) is the video decoder and VCE (video coding engine) is the video encoder. as you can see in the table they are superseded by VCN (video core next)
23:17Venemo: Mangix: it means that the HW decoder doesn't support those things, so they had to be done using shaders
23:19Venemo: you don't have to worry about that on your GPU
23:19Mangix: by video post processing does it mean resizing and shader filters?
23:19Venemo: the article above says exactly what it means
23:19Mangix: oh I see
23:20Venemo: however, that note isn't applicable to your GPU because it is newer
23:20Venemo: you have UVD 3
23:21Venemo: anyway
23:21Venemo: all of this should be supported by Mesa and has been for a long, long time
23:24Mangix: seems to be. I need to tell mpv to use wayland as the backend instead of waylandvk. Otherwise it uses vaapi-copy instead of vaapi
23:25Venemo: dunno what waylandvk is, but if the vk means vulkan, then we already explained that this chip isn't supported by vulkan video
23:25Mangix: yeah. mpv defaults to waylandvk though
23:25KitsuWhooa: it's what mpv uses for video output, not to decode
23:26KitsuWhooa: that's why it probably uses vaapi-copy, because it can't use vaapi with a vulkan output
23:27KitsuWhooa: as a sidenote, I once had an HD 7870XT which I believe is Tahiti and I gave it away back in the day because radeon wouldn't work with it at all and fglrx was.... fglrx.
23:27KitsuWhooa: wish I'd kept it now :p
23:27Mangix: KitsuWhooa: --hwdec can be set to vulkan. I think it's set to auto in my case
23:28KitsuWhooa: that is vulkan video
23:28KitsuWhooa: which is what Venemo mentioned
23:28Mangix: yes
23:28Venemo: it could be someone's pet project to try to enable vulkan video on these. it should be technically possible although of questionable usefulness
23:29Mangix: correct
23:29KitsuWhooa: Only reason I found to use vulkan video over vaapi is that the deinterlacing seemed to be better
23:30Mangix: honestly me playing around with the decode and encode is curiosity. This 3950X is fast enough to not need any of this.
23:31Mangix: *5950X
23:32Venemo: for me the reason was to tick a box
23:32KitsuWhooa: I have a machine I use every now and then with a Xeon X5460 and it very much relies on UVD
23:40Mangix: I mostly care about having a working display which comes with kernel 6.17.9 AFAIK.