#
…

2022-06-05

01:53nuh^ has joined #asahi-re
02:32nuh^ has quit [Ping timeout: 480 seconds]
02:47phiologe has joined #asahi-re
02:50PhilippvK has quit [Ping timeout: 480 seconds]
03:05nuh^ has joined #asahi-re
03:31nuh^ has quit [Ping timeout: 480 seconds]
06:44bisko has joined #asahi-re
06:48bisko has quit []
07:36phirethere's a good chance the new gpu won't require any major changes to the kernel code. Except it will probally come with a new macos version that rearranges all the structs
07:38phireor more, I expect them to use the same agx firmware version across multiple GPU revisions, but that agx firmware might get a version bump
07:57nuh^ has joined #asahi-re
08:09nsklaus has joined #asahi-re
08:23nuh^ has quit [Ping timeout: 480 seconds]
09:22chadmed has quit [Remote host closed the connection]
09:59chadmed has joined #asahi-re
10:08marcanphire: the GPU should be a new major revision AIUI, which will likely make a bunch of changes to initdata and friends
10:09marcanthat's usually how it goes with these things
10:09marcanbut there's no reason for them to change the basic design
10:09marcanso it'll be an ABI bump on that end
10:09marcanI do wonder if they'll keep the structs synced between GPU revisions though
10:10marcanthis is the first time we're going to have two GPU majors supported at the same time
10:11phirehey, maybe they will enable the firmware-managed memory allocations feature
10:12phireI'm not sure if that's an old feature that's disabled on the M1 gpu, or a new feature that's not yet enabled
10:16marcanI want us to support 12.3 for the first GPU driver if possible, mostly to force the issue of supporting multiple firmwares and avoid having to make people upgrade firmware to get GPU support eventually
10:16marcanso that should be interesting if they made big changes
10:17marcanlina might have some work to do there :p
10:17phirepoor lina
10:17marcan(and she has a point with the Rust thing)
10:18phiretechnially we could do a version 0 dirver as just the m1n1 codebase accessing kmem
10:19marcanI'm... not sure that's a good idea, with the UAT and everything involved (and it'd only work for a single app at once)
10:19marcanI've done the "drivers from userspace" thing before and just getting a single buffer mapped to a device was already problematic
10:20phireit certianlly wouldn't be pretty
10:21marcanwhat *might* make more sense for experiments is a really insecure dumb driver that lets you do kernel-space maps and talk to the ASC side
10:21marcanthat would still be single-process but at least it wouldn't be insanity as far as memory management goes
10:21marcanso basically just a UAT driver
10:22marcan(plus the existing rtkit stuff)
10:22marcanalso shouldn't we be having this discussion in -gpu? lina isn't even here
12:56roxfan has quit [Remote host closed the connection]
12:58roxfan has joined #asahi-re
13:12amarioguy has quit [Remote host closed the connection]
13:50nuh^ has joined #asahi-re
14:21the_lanetly_052__ has joined #asahi-re
14:24nuh^ has quit [Ping timeout: 480 seconds]
14:27the_lanetly_052___ has quit [Ping timeout: 480 seconds]
14:57amarioguy has joined #asahi-re
16:52unevenrhombus[m] has quit []
16:52unevenrhombus[m] has joined #asahi-re
16:53unevenrhombus[m] has quit []
16:53unevenrhombus[m] has joined #asahi-re
17:24bisko has joined #asahi-re
17:32bisko has quit [Ping timeout: 480 seconds]
18:01kubes has joined #asahi-re
18:02handofstand[m] has joined #asahi-re
18:14the_lanetly_052__ has quit [Ping timeout: 480 seconds]
19:01kubes has quit [Quit: Leaving]
22:00yrlf has quit [Quit: The Lounge - https://thelounge.chat]
…