#
…

2022-11-08

00:07cylm_ has joined #asahi-gpu
00:11cylm__ has joined #asahi-gpu
00:12cylm___ has joined #asahi-gpu
00:14cylm has quit [Ping timeout: 480 seconds]
00:17cylm has joined #asahi-gpu
00:18cylm_ has quit [Ping timeout: 480 seconds]
00:19cylm_ has joined #asahi-gpu
00:20cylm__ has quit [Ping timeout: 480 seconds]
00:24cylm___ has quit [Ping timeout: 480 seconds]
00:26cylm has quit [Ping timeout: 480 seconds]
01:00chadmed has joined #asahi-gpu
01:20linaalyssa: Aaaa sorry I forgot to push m1n1!!! T_T
01:20linaIt's pushed now...
01:20alyssalina: thanks ^^
01:44nyx_o has joined #asahi-gpu
02:53zzywysm has joined #asahi-gpu
03:29nyx_o has quit [Quit: WeeChat 3.5]
03:35nyx_o has joined #asahi-gpu
06:30cylm_ has quit [Ping timeout: 480 seconds]
06:55bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
07:09chengsun has joined #asahi-gpu
07:12chengsun has quit []
07:18millenialhacker has joined #asahi-gpu
07:43millenialhacker has quit []
07:43bisko has joined #asahi-gpu
07:54bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
08:05chengsun has joined #asahi-gpu
08:08chengsun has quit []
08:14bisko has joined #asahi-gpu
08:15bisko has quit []
08:17SSJ_GZ has joined #asahi-gpu
08:22MajorBiscuit has joined #asahi-gpu
08:36bisko has joined #asahi-gpu
08:37bisko has quit []
08:43bisko has joined #asahi-gpu
08:58bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
09:03bisko has joined #asahi-gpu
09:12bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
09:14bisko has joined #asahi-gpu
09:30bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
10:37ChaosPrincess has quit [Quit: WeeChat 3.6]
10:38ChaosPrincess has joined #asahi-gpu
11:08bisko has joined #asahi-gpu
11:56bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:02bisko has joined #asahi-gpu
12:08bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:08bisko has joined #asahi-gpu
12:14bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:18bisko has joined #asahi-gpu
12:23bisko has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:27bisko has joined #asahi-gpu
12:29bisko has quit []
12:30bisko has joined #asahi-gpu
12:30DarkShadow44 has quit [Quit: ZNC - https://znc.in]
12:30DarkShadow44 has joined #asahi-gpu
13:29alyssaOkay, kernel boots with the new m1n1
13:29alyssaneed to sort out wifi firmware first
13:29alyssaalso my sound broke with the upgrade
13:34alyssaoh, wi-fi firmware Just Works
13:34alyssaso wifi is back and good
13:35alyssalina: is your latest mesa pushed?
13:38alyssaOh my gosh, 3D acceleration! this is awesome!!
13:38alyssalina: i love rust now! ;-D
13:47linaYay!!!! ^^
13:48alyssalina: okay. I love having GPU drivers. haven't touched Rust yet :-p
13:50karolherbstalyssa: hi, how are you today?
13:50linaI still need to fix the allocator to not be totally silly, I'll do that next week...
13:51alyssakarolherbst: I intend to finish MRT before wiring up compute kernels
13:51linaBut I think the driver doesn't do anything silly as far as stability so I hope the kernel side works well for you!
13:51alyssaI like silly
13:51karolherbstalyssa: I was more thinking if you'd like to code some rust, but that's good to know as well :)
13:51linaI think there's like a dozen buffers that get allocated from shmem/mapped/freed for every render w
13:52karolherbstah you want to work on nak instead? that's cool as well!
13:52linawwwwwwwwwwww
13:52alyssastop nerdsniping me :V
13:53lina:lunapat:
13:55alyssa has left #asahi-gpu [#asahi-gpu]
14:11alyssa has joined #asahi-gpu
14:11alyssalina: Debugging the regression you hit whlie rebasing
14:13alyssa1000 commits on main in between, rip
14:15alyssaglmark2 -bdesktop is totally hosed
14:15alyssabut I think that still works on macOS
14:15alyssaso I suspect some UAPI shenanigan
14:17alyssatrying to bisect
14:18alyssabut that's ... nontrivial for obvious reasons
14:22alyssaPreamble shaders did not break things
14:23alyssaAt least, glmark2 -bdesktop still works rebased on top of preamble shaders
14:24alyssaOnce I fix the rebase I'd like to upstream as much of this as possible
14:25alyssagot it down to a range of 481 commits
14:28alyssabut know the guilty commit
14:29alyssaARB_buffer_storage support
14:30alyssalina: Going to need to tag you in on this one
14:30alyssaI can walk you through exactly what's happening in Mesa but I think this is a kernel problem
14:39alyssaokay fixed
14:39alyssanot a kernel problem, modifiers bug
14:40alyssaunsure how any of this worked before tbh
14:40karolherbstI guess you didn't allocate that as linear?
14:41alyssalina: Pushed alyssa/asahi/rebased
14:42alyssasquashed commits because I'm about to do git surgery to upstream as much as possible
15:00alyssaOkay this driver is all sorts of broken :sparkles:
15:00alyssaStuff on the far end of the screen being blurry, that's a fun one
15:00alyssaTexture coordinate precision issue, if I had to guess
15:01alyssaright. Disabling lower_mediump_io "fixes" that
15:01alyssaMaybe an application bug (in sway) but meh
15:10* alyssadoes git surgery
15:12linaalyssa: Thank you!
15:13linaI'm so happy to see you hacking on Linux <3
15:13alyssaIt is what I am here for ;)
15:14linaYou probably saw the UABI version thing, if you have a minute you can actually add the check to make it complain loudly if it mismatches. Then next time I break the UABI it will be obvious if there's a mismatch ^^
15:14alyssaright now I'm just trying to separate out the UAPI stuff from other Mesa changes
15:15alyssaand clean up the other changes and merge them
15:15alyssagive you less to rebase
15:15linaYeah, I think some things might be a bit tricky... you probably also have better ideas of how to integrate it side by side with the macOS stuff
15:15linaThe cmdbuf thing I did is, well... quite nonsense
15:15linaThank you <3
16:37alyssalina: Got bored rebasing and implemented some batch tracking
16:37alyssawhich mitigates the Slow from the transfer_map batch
16:39gtk2 has quit [Remote host closed the connection]
16:39gtk2 has joined #asahi-gpu
16:43alyssacrashing only when I don't use valgrind, that's a classic isn't it
17:09mkurz has joined #asahi-gpu
17:14MajorBiscuit has quit [Ping timeout: 480 seconds]
17:58amarioguy has quit [Remote host closed the connection]
17:58amarioguy has joined #asahi-gpu
18:29Dcow has joined #asahi-gpu
19:32bisko has quit [Ping timeout: 480 seconds]
19:37alyssaAlso trying to figure out why glamor is failing
20:04zshrc has joined #asahi-gpu
20:15alyssagot some splat for you
20:17alyssakodi, which incidentally is totally broken rn
20:17jannauisn't that dcp as agx doesn't use dart?
20:19alyssagot some splat for you
20:19alyssakodi, which incidentally is totally broken rn
20:22zshrc has quit [Ping timeout: 480 seconds]
20:25jannauI guess "WARN_ON(paddr >> cfg->oas)" is very unlikely, what could cause "WARN_ON(pgsize != cfg->pgsize_bitmap)" to trigger? large pages?
20:26svenor something abusing the iommu api
20:26alyssasven: something like my dcp code?
20:26alyssajannau's, sven's, and lina's dcp code
20:26svenlike, say, creating 4K pages, stuffing them into the sg table and then trying to mal that thing
20:26alyssai'm just a userspace person right
20:27svenright, I forgot that you claim that DCP is all jannau these days :P
20:27alyssasven: I don't know anything about that, that's my story and I'm sticking to it :-p
20:28mkurz has quit [Quit: Leaving]
20:40zjstraus has joined #asahi-gpu
20:52zshrc has joined #asahi-gpu
20:56zshrc is now known as Guest898
20:56zshrc has joined #asahi-gpu
21:24steffen[m] has joined #asahi-gpu
21:31zshrc has quit [Quit: leaving]
21:31Guest898 has quit [Quit: leaving]
21:34zshrc has joined #asahi-gpu
21:54alyssalina: Your kernel driver is so much stabler than panfrost.ko, and that's been in prod for years
21:54alyssaMajor kudos
21:54alyssato you and to Rust :-)
22:02alyssaalyssa/asahi/rebased has my rebased
22:06karolherbstthe magic of rust :3
22:07zshrc has quit [Quit: Bye]
22:20nsklaus has joined #asahi-gpu
22:26nsklaus_ has quit [Ping timeout: 480 seconds]
22:27alyssa(which is !19606 + uapi commits)
22:28alyssaTested on t8103, no promises about t6000 but most of today was playing the git rebase game (it's mostly your code, I only rewrote local sections where it made sense) ... so it should still be ok there too
22:29alyssathe git history is mostly nonsense since I wrote all the commit messages but set --author 'Asahi Lina' for the ones that were originally your code ...
22:59maximbazHello, I tried to test the driver on M1 Pro / T6000, which gets only to initializing MMU [1], at which step it crashes on memremap [2] with "Failed to remap handoff region", as per dmesg output [3]. I wonder if I could be doing something wrong, could anyone give me a hint? :)
23:10SSJ_GZ has quit [Ping timeout: 480 seconds]
23:24jannaumaximbaz: does the device have 32GB RAM? is there a warning in dmesg?
23:39cr1901 has quit [Read error: Connection reset by peer]
23:46cr1901 has joined #asahi-gpu
…