#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 13:10
karolherbstalso for the GPU shutdown hack: can you test if doing only a WFI on both sides (GPU and ASC) also solves the issue? The kernel driver overwriting/changing state while it's still in use by other pieces still sounds like the most likely problem here to me.#asahi-gpu · 2022-10-13 13:06
karolherbstlina: having the same VA on both sides is something which is also required once you are doing SVM like things. So totally not surprised Apple is doing it, because it helps a lot if you share deep data structures (think linked lists) between both sides a lot. I suspect their design makes this much easier, but there might also be a flag to not do this.#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 21:43
alyssathat sounds dcp#asahi-gpu · 2022-10-12 21:43
jannaumaybe it tries to select the mode with the highest refresh rate and you capture card additional unusual modes besides the 64Hz FullHD mode#asahi-gpu · 2022-10-12 21:41
jannaunot sure why glmark2 hits that. For some reason it must chose a different mode than all other software#asahi-gpu · 2022-10-12 21:40
jannaustarnge that dcp reports timing modes without color modes#asahi-gpu · 2022-10-12 21:40
jannaulina: https://github.com/jannau/linux/commit/60babbde481e945a979b2c1c368d93d7cbdce1f5 should fix the dcp issue. newer tree than what you merged but that commit should be cherry-pickable#asahi-gpu · 2022-10-12 21:36
alyssaquery sven#asahi-gpu · 2022-10-12 21:36
alyssahello fruit farmer#asahi-gpu · 2022-10-12 21:29
mactroon66[m]they should have adopted c++ instead of the tranny sjw language#asahi-gpu · 2022-10-12 21:29
mactroon66[m]rust is the biggest disgrace that happened to linux kernl#asahi-gpu · 2022-10-12 21:26
mactroon66[m]have you dilated your stinkditches today yet#asahi-gpu · 2022-10-12 21:25
mactroon66[m]hello trannies#asahi-gpu · 2022-10-12 19:59
alyssalol#asahi-gpu · 2022-10-12 19:29
svennerd sniping successful :>#asahi-gpu · 2022-10-12 19:28
alyssao:)#asahi-gpu · 2022-10-12 19:28
alyssayou're the dcp maintainer now, it's your call :p#asahi-gpu · 2022-10-12 19:28
alyssashrugs#asahi-gpu · 2022-10-12 19:21
jannauon the other hand judging by the amount of changes they did between macos 11 and 12 it should be manageable in C#asahi-gpu · 2022-10-12 19:21
jannauI won't mind Rust and depending on how much Apple changed the DCP rpc in ventura I think it could be useful#asahi-gpu · 2022-10-12 19:19
alyssai don't#asahi-gpu · 2022-10-12 19:19
svenah, don't really care. if you want to have dcp rust I'll gladly learn it to implement the EPIC crap there *shrug*#asahi-gpu · 2022-10-12 19:17
alyssaacronym#asahi-gpu · 2022-10-12 19:16
alyssaRiiR is a very important algorithm for being friends with lina~ ;-p#asahi-gpu · 2022-10-12 19:16
svenmeh, i'll take care of that later this week#asahi-gpu · 2022-10-12 19:15
svenyup, because I also used tps6598x as template :D#asahi-gpu · 2022-10-12 19:15
jannaure-implemented in rust#asahi-gpu · 2022-10-12 19:15
jannaulol, with a clash in the filename? I looked at tps6598x as template#asahi-gpu · 2022-10-12 19:15
svenRiiR?#asahi-gpu · 2022-10-12 19:14
svenoh well..#asahi-gpu · 2022-10-12 19:14
alyssaI'd rather it weren't but otoh maybe the mode parser should be.#asahi-gpu · 2022-10-12 19:14
svenlol, i also have a commit in my tree to move some dev_dbg to tracepoints#asahi-gpu · 2022-10-12 19:13
alyssais dcp getting RiiR'd or not I can't remember#asahi-gpu · 2022-10-12 19:13
svenlet me see if I can at least still rebase#asahi-gpu · 2022-10-12 19:12
svenand i have some bastard tree forked off a few commits ago that implement yet another dumb parser for these EPIC things :(#asahi-gpu · 2022-10-12 19:12
jannaunot yet what is used in the agx tree#asahi-gpu · 2022-10-12 19:12
alyssaseems legit#asahi-gpu · 2022-10-12 19:12
jannaucurrent tree is https://github.com/jannau/linux/tree/asahi-dcp-october#asahi-gpu · 2022-10-12 19:11
jannaudcpep.c is now iomfb.c btw#asahi-gpu · 2022-10-12 19:10
jannauI'll fix it and look at the WARN_ONs#asahi-gpu · 2022-10-12 19:10
alyssaone of those needs to change, idk what colour the bikeshed should be#asahi-gpu · 2022-10-12 19:10
alyssabut dcpep.c assumes 0 means there's a valid mode#asahi-gpu · 2022-10-12 19:09
alyssaactual bug though: if no modes match, the mode parser returns 0 (success) with best mode = -1#asahi-gpu · 2022-10-12 19:08
alyssathis one is bogus#asahi-gpu · 2022-10-12 19:07
alyssa WARN_ON(endpoint != DCP_ENDPOINT);#asahi-gpu · 2022-10-12 19:07
alyssathese are bogus assertions that should be dev_err instead#asahi-gpu · 2022-10-12 19:07
alyssa WARN(!dcp->connector->connected, "can't flush if disconnected")) {#asahi-gpu · 2022-10-12 19:07
alyssa if (WARN(dcp_channel_busy(&dcp->ch_cmd), "unexpected busy channel") ||