#asahi-gpu · 2022-10-13 15:10
linaBecause previously I could reproduce the problem with kmscube (at least sometimes) and now I can't...#asahi-gpu · 2022-10-13 15:08
linaI'm already doing that, all submissions are synchronous right now... but I've been messing with stuff today and now I'm wondering if I'm confusing myself again.#asahi-gpu · 2022-10-13 15:01
linaI'm not sure what you mean with doing a WFI. I can't change ASC code so I can't force it to do that, and what does WFI mean for the GPU?#asahi-gpu · 2022-10-13 15:00
linaWell... sure, but that's the same as just using 40-bit VAs to begin with, which isn't great...#asahi-gpu · 2022-10-13 14:59
linaIt's like the GPU team took a few steps in that direction but they're not there yet?#asahi-gpu · 2022-10-13 14:59
linakarolherbst: Except for the part where GPU VAs are only 40 bits so it's incompatible with more common 48-bit VA configs on the AP side... and the firmware doesn't support demand paging either. Apple don't use 1:1 mappings at all.#asahi-gpu · 2022-10-13 08:15
linaI wrote up what I know so far about AGX caching/MMUs/TLBs and the issues I'm running into, in case anyone's interested: https://gist.github.com/asahilina/1a70f1f26b9cd3c52d61d2b8e8b27673#asahi-gpu · 2022-10-12 17:28
linaFor the coprocessor I'm just never unmapping/freeing anything for now, to eliminate variables (besides, I plan to do that in my allocator later on anyway, there's no good reason to thrash those mappings)#asahi-gpu · 2022-10-12 17:27
linaBut there's something else wonky going on with the GPU. If I force it to stay on on purpose, sometimes it crashes on the second frame, sometimes a bit later.#asahi-gpu · 2022-10-12 17:27
linaI might still have a TLB issue with the coprocessor, which was being hidden by the GPU shutdown thing beacuse I think the coprocessor shuts down shortly after the GPU (so it mostly worked in practice)#asahi-gpu · 2022-10-12 17:26
linaI actually don't think I have a TLB issue at this point, at least not with the GPU. As far as I can tell from the m1n1 experiment, TLBI actually works fine#asahi-gpu · 2022-10-12 17:25
linaI doubt Apple's cores have Cortex errata... ^^;#asahi-gpu · 2022-10-12 17:05
linaNow it crashes even more! (Which is probably a good thing for debugging)#asahi-gpu · 2022-10-12 17:04
linaNope!#asahi-gpu · 2022-10-12 17:02
linaI think it was drawing 512x512 squares for the cursor or something like that ^^#asahi-gpu · 2022-10-12 16:42
linaglmark2 makes it go explodey#asahi-gpu · 2022-10-12 16:39
linaI think it's your problem, but it's not a blocker for what I'm trying to do ^^#asahi-gpu · 2022-10-06 14:22
linaI don't remember how much initdata changed from 12.1 to 12.3, but I think it wasn't zero#asahi-gpu · 2022-10-06 14:22
linaOh...#asahi-gpu · 2022-10-06 14:21
linaYou've been testing on 12.1 or 12.6?#asahi-gpu · 2022-10-06 14:20
linaNo, it's 12.3#asahi-gpu · 2022-10-06 13:55
linaphire: Wait, you're on 12.6? The supported version is 12.3 right now...#asahi-gpu · 2022-10-06 08:31
linaI wonder if it would match better at another version...#asahi-gpu · 2022-10-06 08:30
linaphire: Maybe the firmware versions actually diverge? Like G13X got forked off of G13G before G14G support got added or started to be added?#asahi-gpu · 2022-10-06 08:29
linaOn macOS, yes (on Linux I do it in the kernel right now)#asahi-gpu · 2022-10-06 06:06
linaAnd that would cause breakage only when preemption happens, so "flakiness"#asahi-gpu · 2022-10-06 06:06
linaIt makes sense that those would have to be larger for more cores#asahi-gpu · 2022-10-06 06:06
linaRandom thought: maybe the t6k broken mesa side on macOS was due to too-small preemption buffers?#asahi-gpu · 2022-10-05 23:49
linajannau: The person testing on the Air also reported that quitting sway causes the brightness to step down every time, until it goes to 0#asahi-gpu · 2022-10-05 23:48
linaphire: Conjecture: there is a union of HW-related data for different chips, that's why it resyncs#asahi-gpu · 2022-10-05 23:47
linaOr maybe it's the max number of physical cores, not actually enabled ones?#asahi-gpu · 2022-10-05 23:47
linaphire: Someone is testing on a 7-core Air and it seems to work, not sure if the driver is expected to get/forward the number of cores somewhere?#asahi-gpu · 2022-10-05 23:45
linajannau: Yeah, there is no fault handling yet. Right now you have to track down those faults manually with the hypervisor.#asahi-gpu · 2022-10-05 23:44
linaphire: I think I saw that already in t8103, but I have no idea what it is...#asahi-gpu · 2022-10-05 15:11
linaWeston needs it too, yes!#asahi-gpu · 2022-10-05 15:08
linaOtherwise there is some dbus timeout thing#asahi-gpu · 2022-10-05 15:08
linaNo, if you're running gnome-session manually, prefix it with dbus-run-session#asahi-gpu · 2022-10-05 15:06
linaSomeone's already running Minecraft I hear#asahi-gpu · 2022-10-05 14:42
linaThank you!#asahi-gpu · 2022-10-04 05:55
linaThere will be a demo tonight ^^#asahi-gpu · 2022-10-03 08:12
linaMaybe even less, hard to tell exactly with the firmware power management time slice#asahi-gpu · 2022-10-03 08:12
linaYeah, the GPU turns on and off in less than 2ms#asahi-gpu · 2022-10-02 16:53
linaYup...#asahi-gpu · 2022-10-02 16:52
linaalyssa: Yeah ^^#asahi-gpu · 2022-10-02 16:52
linaThe stories of Intel internal coporate culture explain so much...#asahi-gpu · 2022-10-02 16:51
linaI read until the leaving intel part#asahi-gpu · 2022-10-02 16:51
linaAlso it seems so obvious in retrospect, shaders for CPUs... why wasn't that a thing before?#asahi-gpu · 2022-10-02 16:50
linaIt is!#asahi-gpu · 2022-10-02 16:45
linaI ended up reading the entire ispc saga...