#asahi-dev · 2026-07-20 13:53
sofuswith VV (right now at least) we would also be "responsible" for parsing bitstream stuff#asahi-dev · 2026-07-20 13:52
sofusif we do v4l2, its only submitting parameters to the HW we are "responsible for"#asahi-dev · 2026-07-20 13:52
sofusto me the question of VV vs v4l2 is kinda more where we put code, and who is responsible#asahi-dev · 2026-07-20 13:50
sofusndufresne: nice, thanks! It looks our driver will be the "most slicebased one"#asahi-dev · 2026-07-20 13:33
sofusbut it would be slower and more error prone#asahi-dev · 2026-07-20 13:32
sofusyeah, in theory we could let the cm3 do everything#asahi-dev · 2026-07-20 13:13
sofusi think we should do VV if we dont do v4l2#asahi-dev · 2026-07-20 13:02
sofusbut i think the work required would be the same#asahi-dev · 2026-07-20 13:01
sofusvaapi is also missing a few things, but its not as bad#asahi-dev · 2026-07-20 12:56
sofusno that makes sense, avd is very much slice based#asahi-dev · 2026-07-20 12:54
sofusi looked at the headers here: https://gitlab.freedesktop.org/mesa/mesa/-/tree/main/include/vk_video#asahi-dev · 2026-07-20 12:52
sofusleio: unless i missed something its not enough (its the same as vaapi)#asahi-dev · 2026-07-20 12:38
sofusleio: am i missing something? It looks like the bindings are only for nouveau?#asahi-dev · 2026-07-20 12:32
sofusmaybe im missing something, this is mainly for the slices/tiles#asahi-dev · 2026-07-20 12:28
sofus(we would need to do this for vaapi/vk translation layer anyway)#asahi-dev · 2026-07-20 12:28
sofusvk video does not seem to expose all the things we need. We would need to parse the bitstream #asahi-dev · 2026-07-20 11:39
sofusie. we would not have to make it ourselfs#asahi-dev · 2026-07-20 11:39
sofusi would rather make a vk translation layer as i suspect other people would be interested in that#asahi-dev · 2026-07-20 11:37
sofusfrom the little i tested the vaapi brigde thing it did not seem like it used much more power#asahi-dev · 2026-07-20 11:34
sofusim kinda thinking i should fix the og one#asahi-dev · 2026-07-20 11:33
sofusleio: yes, although i kinda suspect its at least partially slop#asahi-dev · 2026-07-20 11:29
sofus... rust would have been nice for this#asahi-dev · 2026-07-20 11:29
sofusis there a reason i should not fill the variant structs at probe time?#asahi-dev · 2026-07-20 11:26
sofuschadmed: will do my best to make it generic, although based on changes between rev3 and rev4 im not really worried#asahi-dev · 2026-07-20 11:11
sofusbut like the differences between variants are at most 3 if statments, i dont think thats enough for a seperate function#asahi-dev · 2026-07-20 11:10
sofusah, you are thinking about the addresses right? i would rather do ops->set_addr for that#asahi-dev · 2026-07-20 11:04
sofusmaybe making shared functions for the common stuff is simpler#asahi-dev · 2026-07-20 10:58
sofusthe instruction sequence looks kinda like hdr -> codec_flags -> codec_bufs -> output -> ref_buffers -> slices/tiles#asahi-dev · 2026-07-20 10:57
sofusre the avd rewrite, im also debating if i should make the instruction generation into a shared function and use function tables for the codec specifik things#asahi-dev · 2026-07-20 10:29
sofusyeah thats kinda the same as avd rev3 with its extra 0 commands#asahi-dev · 2026-07-20 10:26
sofusJamesCalligeros[m]: i meant mostly in regards to the macros, but i suppose its fine#asahi-dev · 2026-07-20 10:18
sofusWe cant really get rid of this because of devices with never versions right? Whats the plan to upstream this? (if there is one)#asahi-dev · 2026-07-20 10:15
sofusthats cool! This is only for for generating the iomfb things right?#asahi-dev · 2026-07-20 10:03
sofusi dont really get how dcp's macros work, like how does it know what version to build with?#asahi-dev · 2026-07-20 10:01
sofusso like defining everything in something like https://github.com/AsahiLinux/linux/blob/f4dd286f7888b348c757b9a2f28dd7bde4c3532b/drivers/media/platform/apple/avd/avd-hevc.c#L448#asahi-dev · 2026-07-20 10:01
sofusthis is mostly for the instruction stream definitions#asahi-dev · 2026-07-20 10:01
sofusthankfully i dont know of any differences#asahi-dev · 2026-07-20 09:51
sofus200 is a guess, if i dont make them shared it will be around 400#asahi-dev · 2026-07-20 09:50
sofusdifferent things, and they are shared between codecs, but have slighly different meanings#asahi-dev · 2026-07-20 09:47
sofusim also bad at names, so having to define more than 200 things kinda scares me :(#asahi-dev · 2026-07-20 09:45
sofushmm, im trying to plan how to "prettify" the avd driver. rkvdec's approach seem unfeasable for avd, if anyone knows a better approach than define+genmask+field_prep (for register like things) please let me know!#asahi-dev · 2026-07-17 18:39
sofus708 on t6020#asahi-dev · 2026-07-17 08:37
sofusah i had forgotten#asahi-dev · 2026-07-17 08:32
sofusi like that it formalizes both the sequence and bit layout#asahi-dev · 2026-07-17 08:32
sofusfor cleaning up avd, maybe something like this would be nice? https://github.com/AsahiLinux/linux/blob/f4dd286f7888b348c757b9a2f28dd7bde4c3532b/drivers/media/platform/rockchip/rkvdec/rkvdec-vdpu381-regs.h#L46#asahi-dev · 2026-07-17 06:53
sofusyuka: wow thats really fast!#asahi-dev · 2026-07-17 06:19
sofuschadmed: nice! I have no idea how much power is expected for a heatpipe. For me its always ~30W when doing the stress test and sits idle att 1-2 W#asahi-dev · 2026-07-16 19:18
sofuswhops#asahi-dev · 2026-07-16 19:17
sofus:q#asahi-dev · 2026-07-16 16:25
sofusalso when running something like stress-ng it goes up to ~30W, that seems like a lot