#asahi-dev · 2026-06-29 08:11
chadmedand 313#asahi-dev · 2026-06-29 08:10
chadmedi set the indexes to 1 accidentally for j413#asahi-dev · 2026-06-29 08:10
chadmedah fuck this one's my fault#asahi-dev · 2026-06-29 08:08
chadmedah okay yeah its the index thing then i think#asahi-dev · 2026-06-29 08:08
integralpilotim trying it now#asahi-dev · 2026-06-29 08:08
integralpilot*thing#asahi-dev · 2026-06-29 08:07
integralpilotso it might be the 1 to 2 think#asahi-dev · 2026-06-29 08:07
integralpiloti think jannau pushed v7.1 with the corrected/swapped masks for j613#asahi-dev · 2026-06-29 08:07
integralpiloti think the masks were right, if i swap them to what you said (0 0xf0f0) it becomes 0 and near max int size#asahi-dev · 2026-06-29 08:07
integralpilotah :P#asahi-dev · 2026-06-29 08:07
chadmedindex 1 is the left tweeter, so the current setup just idles out every left tdm slot :P #asahi-dev · 2026-06-29 08:06
chadmedalso swap the masks, 0 should be 0xf0f0 and 2 should be 0x0f0f... i think #asahi-dev · 2026-06-29 08:06
chadmeddai-tdm-slot-tx-idle-mask-2 = <0xf0f0>;#asahi-dev · 2026-06-29 08:06
chadmeddai-tdm-idle-mode-2 = "zero";#asahi-dev · 2026-06-29 08:05
chadmedchange idle mode and mask 1 to 2 lol#asahi-dev · 2026-06-29 08:05
integralpilotnot sure what changed in v7.1 now so that it doesnt work anymore :(#asahi-dev · 2026-06-29 08:05
integralpilotnope swapping them didnt help this time#asahi-dev · 2026-06-29 08:00
integralpilothaha no worries#asahi-dev · 2026-06-29 08:00
chadmedsorry long day at the bullshit factory#asahi-dev · 2026-06-29 08:00
chadmedoh thats what we swapped to fix it yeah thats right#asahi-dev · 2026-06-29 08:00
integralpilotbut aren't working so ill try swapping them#asahi-dev · 2026-06-29 08:00
integralpiloti meant the bus keeper masks are in the new spot on j613#asahi-dev · 2026-06-29 07:59
integralpilotoh no i got rid of the swapped v/i slots yesterday as i didnt think that was right#asahi-dev · 2026-06-29 07:59
chadmedyour swapped v/i slots?#asahi-dev · 2026-06-29 07:59
integralpilotthey just must be wrong, i'll try swapping them#asahi-dev · 2026-06-29 07:59
integralpilotyep they are there#asahi-dev · 2026-06-29 07:55
chadmedhave a look at e.g. j415 #asahi-dev · 2026-06-29 07:55
chadmedthey go on the dai link now#asahi-dev · 2026-06-29 07:54
chadmedare the bus keeper masks in the right place#asahi-dev · 2026-06-29 07:54
integralpiloti will try and find the new version of ti,stdout-force-zero-mask as i think you said that changed with v7.1#asahi-dev · 2026-06-29 07:54
integralpilotjannau: with the new asahi-wip v7.1, mba internal speakers dont work anymore (speakersafetyd gets wrong data and panics)#asahi-dev · 2026-06-29 07:48
sofuswhere telling avd where the destination is, and a bunch of other things, some of the instructions go directly to pp#asahi-dev · 2026-06-29 07:45
sofusits not only for programming the vp#asahi-dev · 2026-06-29 07:45
sofusah, you mean what the stream_hdr does. We need to do that as far as im aware #asahi-dev · 2026-06-29 07:43
chadmedor am i misunderstanding something#asahi-dev · 2026-06-29 07:43
chadmedthis is fine, but seems like a waste of the AP's time to do this for every frame#asahi-dev · 2026-06-29 07:43
chadmedthe way i understand it now, every time we get a request to decode a frame, we reprogram the entire VP with the stream's parameters regardless of whether or not it was already decoding frames from that stream#asahi-dev · 2026-06-29 07:42
sofuschadmed: i actually wanted clarification on what you mean by that#asahi-dev · 2026-06-29 07:41
chadmedand see if we could possibly make use of multiple vps, but if not even apple does it then its a nothing problem #asahi-dev · 2026-06-29 07:41
chadmedi just wanted to avoid the "program the stream for every single slice" thing but w/e #asahi-dev · 2026-06-29 07:41
chadmedoh ok who cares then#asahi-dev · 2026-06-29 07:40
jannauplease don't. the APIs we want to support (va-api/vulkan video) are stateless as well#asahi-dev · 2026-06-29 07:40
chadmedi just dont have any confidence in the "simply wait for literally everything to properly support v4l2 stateless or fix all of userspace ourselves" route #asahi-dev · 2026-06-29 07:39
chadmedthat was partially what i was thinking and annoyingly we may have to, at least partially#asahi-dev · 2026-06-29 07:39
sofuschadmed: you could go the crazy route and use the cm3 to make avd statefull#asahi-dev · 2026-06-29 07:39
jannauI think I remember a toot mentioning speakersafetyd and iirc ayn devices#asahi-dev · 2026-06-29 07:38
chadmedand i want it to be easy to support not-v4l2request apis#asahi-dev · 2026-06-29 07:38
sofusand btw, i have never gotten apple to use more than one vp#asahi-dev · 2026-06-29 07:38
jannaumac mini/studio don't have "sound-name-prefix", easily changed#asahi-dev · 2026-06-29 07:37
chadmedagreed, im just having fun + playing