#
✕ nick:jannau
#asahi-gpu · 2025-07-23 15:58
jannaucurrently running VK cts on that branch
#asahi-gpu · 2025-07-23 15:52
jannaualyssa: any concerns over tagging https://gitlab.freedesktop.org/jannau/mesa/-/commits/asahi-20250723-cand?ref_type=heads as asahi-20250723? FAR mesa is still at ~25.1.0-rc2
#asahi-gpu · 2025-07-20 08:55
jannaubanned LLM!ircbot@* let's hope that holds
#asahi-gpu · 2025-07-11 15:43
jannauI was careless when I adapted it to upstreams set_len() -> {dec,inc}_len change. Added set_len() back for now
#asahi-gpu · 2025-07-11 15:27
jannauyes and duplicating items
#asahi-gpu · 2025-07-11 15:08
jannaugpu on 6.15 fixed. wasn't quite what I suspected but my nemesis: Vec::drain()
#asahi-gpu · 2025-07-09 15:11
jannauexpect maybe there's something dropped from gem/gem::shmem
#asahi-gpu · 2025-07-09 15:11
jannaunot quite sure how I'm losing the fences. I don't see any changewhich affects them directly
#asahi-gpu · 2025-07-09 15:08
jannauand the pin-init move to it's own crate
#asahi-gpu · 2025-07-09 15:08
jannauI pulled patches from v6.16-rc1 and updates from the mailing list. that includes the upstream drm abstraction and reworked gem shmem and the upstream xarray and a bunch of minor other stuff
#asahi-gpu · 2025-07-09 14:54
jannaufinally some rendering on 6.15. only a few frames though since syncobj/fences are somehow broken / getting dropped
#asahi-gpu · 2025-06-25 06:42
jannauI've created a nightly build for that: https://copr.fedorainfracloud.org/coprs/g/asahi/mesa-nightly/build/9205574/
#asahi-gpu · 2025-06-16 20:05
jannauhttps://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35562
#asahi-gpu · 2025-06-16 20:03
jannaulol, we inadvertently sabotaged Xorg on asahi since summer last year
#asahi-gpu · 2025-05-26 19:52
jannauESR 0x96000004, level 0 translation fault while reading 0xd6580ffffffa8
#asahi-gpu · 2025-05-26 19:50
jannaufw crash on the first repeat
#asahi-gpu · 2025-05-26 19:34
jannaufirst VK cts on asahi-6.13.8-1 + dts fix + mesa asahi-20250221 finished without GPU fw crash and only a single "** GPU timeout nya~!!!!! **" at the end
#asahi-gpu · 2025-05-26 19:13
jannaudoes the CTS cache shaders? the old CTS build works again after downgrading mesa to asahi-20250221
#asahi-gpu · 2025-05-26 18:47
jannauthe apple,ppm-ki "<30>" vs. "<30.0>" typo is there since asahi-6.1-1 (likely the beginning of GPU support)
#asahi-gpu · 2025-05-26 18:41
jannauI think it was an issue from when I started running CTS on the studio
#asahi-gpu · 2025-05-26 18:40
jannauhad to use --jobs 6 or less to get full CTS runs. I saw that before september both with GL and VK CTS
#asahi-gpu · 2025-05-26 18:37
jannauit easily reproduced on g13d. I haven't tested that much on g13c
#asahi-gpu · 2025-05-26 18:26
jannauack. little annoying that it makes testing the devicetree changes harder. I guess I could use old kernel and mesa
#asahi-gpu · 2025-05-26 16:34
jannauthe alternative would be using reserved-memory for the initdata
#asahi-gpu · 2025-05-26 16:33
jannauwas robher ok with stuffing 27kb + 6kb + 92kb (or so for the Ultras) of binary data into the dtb? I find that at least inconvenient since I occasionally look at the decompiled devicetree from /proc/device-tree
#asahi-gpu · 2025-05-26 16:30
jannauwe could test whether the asahi/hk enable PR crashes the firmware. if it does bisecting would be easy
#asahi-gpu · 2025-05-26 16:22
jannauok, that at least confirms that's not the devicetree/initdata fixes
#asahi-gpu · 2025-05-26 16:18
jannauare you testing with the stock fedora kernel or slef-built asahi-wip?
#asahi-gpu · 2025-05-26 16:17
jannauwhile or after the changes?
#asahi-gpu · 2025-05-26 16:13
jannaubisecting would be annoying over the uapi changes
#asahi-gpu · 2025-05-26 16:12
jannaualyssa: maybe the our kernel submit "regression" made it unlikely to hit. maybe you fixed something while reworking the uapi
#asahi-gpu · 2025-05-23 16:32
jannaualyssa: are you still seeing firmware crashes during CTS runs? I haven't seen one sofar today while running deqp-runner with 10 jobs on a 10-core m2 pro
#asahi-gpu · 2025-05-23 11:37
jannaunot great that the old build can make the kernel driver/gpu non-functional and not clear why that happens at all
#asahi-gpu · 2025-05-23 11:35
jannauwith the new CTS build mesa/main and the system drivers work
#asahi-gpu · 2025-05-23 07:25
jannauI'll redo the test with the system to see if it was the old cts build (not sure how though)
#asahi-gpu · 2025-05-23 07:24
jannauI'v updated CTS to vulkan-cts-1.4.2.1 and rebuilt (still missing cstdint includes) and currently testing mesa main. that seems to work
#asahi-gpu · 2025-05-23 07:22
jannauthat was with a vulkan-cts-1.4.1.1 CTS build built on fedora41
#asahi-gpu · 2025-05-23 07:21
jannaualyssa: issue reproduces on m2 pro mac mini as well https://gist.github.com/jannau/fb0ea4a1333302e0476951315fe4b58f
#asahi-gpu · 2025-05-22 20:27
jannaualyssa: have you done vulkan CTS runs on f42 with mesa-asahi-20250425? I'm running into blocking GPU timeouts after 3-4 minutes. It's also glacially slow (5h estimated remaining)
#asahi-gpu · 2025-04-28 13:27
jannauperf annotations looked like submit is spinning on something. I need to redo that with the copr build kernel to get debug symbols to work
#asahi-gpu · 2025-04-24 19:07
jannautagged the kernel with the uapi, expecting bug reports any minute
#asahi-gpu · 2025-04-24 18:54
jannaunew build underway
#asahi-gpu · 2025-04-24 18:07
jannaudoes it make sense to test if it fixes parallel CTS runs
#asahi-gpu · 2025-04-24 18:06
jannaualyssa: I have builds in my uapi copr but I wouldn't say too late
#asahi-gpu · 2025-04-18 18:59
jannaunot sure if that's an issue in other VMMs than muvm. for muvm we can either ship patches or just not build venus
#asahi-gpu · 2025-04-18 18:58
jannauI have fixes for the venus issue https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34613
#asahi-gpu · 2025-04-18 15:02
jannaumy local is a random mesa checkout from yesterday + virtgpu + my fixes but rpm build uses 25.1.0-rc1 + patches
#asahi-gpu · 2025-04-18 15:00
jannauyes, it's just not in 25.1.0-rc1
#asahi-gpu · 2025-04-18 14:44
jannaucommit 9b55451ea7b ("hk: fix underbinding scratch")
#asahi-gpu · 2025-04-18 14:42
jannauhuh, a VM_BIND with range=0x400 in 25.1-rc1 I don't see with 6d70ec449ff (random commit from yesterday)

plus ›