#asahi-dev · 2026-10-16 09:23
sofusis 'IX' the prefered abbreviation for interchange? Looks cooler than my 'IC'#asahi-dev · 2026-10-15 12:36
sofuschadmed: thanks!#asahi-dev · 2026-10-15 11:36
sofusthanks, i wonder how this will work with linux-firmware#asahi-dev · 2026-10-15 11:34
sofusjannau: does it matter what the default is?#asahi-dev · 2026-10-15 11:29
sofuschadmed: feel free to add and fix it. Otherwise i will do a pr later#asahi-dev · 2026-10-15 11:22
sofuschadmed: is there a reason for not including https://github.com/sofus13/linux/commit/0890ff143896ef9226e8ab2b4d56fc42b3a8b4f7#asahi-dev · 2026-10-07 18:38
sofusjannau: thanks! Will watch later. I did have VV in mind when designing the driver, so porting should be straigtfull#asahi-dev · 2026-10-06 09:18
sofusoh wait sorry#asahi-dev · 2026-10-06 09:18
sofusnot really tho, it would also need to do the same in avd-drv#asahi-dev · 2026-10-06 09:16
sofusits done totally differently in my ml version. I can update the whole thing to use kvzalloc instead but thats likely not necessary if 615 gets merged#asahi-dev · 2026-10-06 09:02
sofuschadmed: i think https://github.com/AsahiLinux/linux/pull/615 would fix it#asahi-dev · 2026-10-06 08:59
sofushuh, i could open it#asahi-dev · 2026-10-04 12:09
sofuschadmed: description is still correct, the issue was bpp#asahi-dev · 2026-10-03 09:57
sofusbut there are still the same suspicious writes#asahi-dev · 2026-10-03 09:56
sofushmm, on my m3 it has not rebooted yet, i waited ~9min#asahi-dev · 2026-10-03 09:21
sofuswhen you tried getting macos to suspend?#asahi-dev · 2026-10-03 08:58
sofusp.write32(u.adt["arm-io/pmgr"].get_reg(38)[0] + 0x4000, 1)#asahi-dev · 2026-10-03 08:58
sofusno ~50 seconds#asahi-dev · 2026-10-03 08:23
sofusi wonder what cancels that#asahi-dev · 2026-10-03 08:19
sofusmasking that should at least be usefull for figuring out resume stuff right?#asahi-dev · 2026-10-03 08:19
sofusit seems macos starts some kinda reboot timer right before suspending. Its just a random pmgr write, writing 0 does not cancel it#asahi-dev · 2026-10-01 11:11
sofusyeah that lookes fine#asahi-dev · 2026-10-01 11:02
sofusthe first one was just the typical red square#asahi-dev · 2026-10-01 10:46
sofusit expects rgb yeah, but it does not look completly broken#asahi-dev · 2026-10-01 10:31
sofusi tried changing it, but there was no difference (which i thought was confusing)#asahi-dev · 2026-10-01 10:27
sofuschadmed: yes i have the patched mesa, tile size seems to be wrong for luma#asahi-dev · 2026-09-30 13:46
sofusbut it only works when it does not touch the gpu#asahi-dev · 2026-09-30 13:44
sofusi got the vaapi wrapper to use interchange, i had trouble last time because my compositor did not use vulkan#asahi-dev · 2026-09-26 13:20
sofusjust sendt v2 :) Its actually ~3 seconds faster on all fluster tests#asahi-dev · 2026-09-25 12:56
sofuswe cant use 'drm_format_info_bpp' then...#asahi-dev · 2026-09-25 12:50
sofusthanks! That makes sense, now i also get how to read it. Thats p030 then#asahi-dev · 2026-09-25 12:42
sofusi dont know how to check. I could only test with interchange#asahi-dev · 2026-09-25 12:37
sofusi was unsure if NV15 translates to dcp format P420#asahi-dev · 2026-09-25 12:37
sofusi confused myself then. nv15 gives the correct bpp for the interchange buffer calculations, i dont know if avd can output nv15 withouth interchange#asahi-dev · 2026-09-25 12:33
sofusjannau: i might have confused myself. Is that nv15 or p420 you are talking about? 4 values in 5 bytes is nv15#asahi-dev · 2026-09-25 11:56
sofusthe definition did not make me a lot smarter. But nv15 gives the correct bpp#asahi-dev · 2026-09-25 11:49
sofusavd v2 is gonna have quite a lot of changes, not sure if i should also include the multi-core stuff#asahi-dev · 2026-09-25 11:45
sofusJamesCalligeros[m]: lol, im not actually sure its nv15/20/30 but the math is mathing. the 'p' could be for packed#asahi-dev · 2026-09-25 11:42
sofusnot only no, avd's interchange uses the new formats i added#asahi-dev · 2026-09-25 11:39
sofusand my mystery format is actually just p030#asahi-dev · 2026-09-25 11:38
sofusbut like p010 did not look completly wrong so i kinda wonder if avd could also use the 'padded' formats for interchange#asahi-dev · 2026-09-25 11:34
sofuschadmed: for interchange https://github.com/sofus13/linux/tree/dcp/interchange#asahi-dev · 2026-09-24 11:35
sofusi also only learned about it for testing this specifik feature. I wish i can get it to run on wayland#asahi-dev · 2026-09-24 11:29
sofusbut anyway, my main goal was just testing everything works#asahi-dev · 2026-09-24 11:28
sofusndufresne: i saw the HW specifik calculations but im not sure they make a difference when using the kmssink#asahi-dev · 2026-09-24 10:18
sofuslol yeah 'KMS dumb buffers are not suitable for hardware-accelerated rendering nor video decoding'#asahi-dev · 2026-09-24 10:05
sofusit gets it via DRM_IOCTL_MODE_CREATE_DUMB#asahi-dev · 2026-09-24 10:00
sofusi found it, its using the pitch 'calculcated by the kms driver' to calculate the offsets according to a comment#asahi-dev · 2026-09-24 09:48
sofusi calculate the offset in the dcp driver right now, but i would assume thats forbidden#asahi-dev · 2026-09-24 09:46
sofusi think thats for something else. At least with gstreamer i think it ignores all v4l2 fields for the drm sizes/pitches