#asahi-dev · 2026-08-02 10:25
chadmeddoes it being an easy diff vs some other vendor driver actually bring us any benefits at all though#asahi-dev · 2026-08-02 10:25
sofusc file is not as clean, but still https://pastebin.com/b26uMiz9#asahi-dev · 2026-08-02 10:21
sofushttps://pastebin.com/THkZsndq#asahi-dev · 2026-08-02 10:21
sofuswhile i agree on the style things, im not sure if it makes sense to change the verisilicon copy paste, its very easy to diff right now#asahi-dev · 2026-08-02 10:07
chadmedsofus: left a small review, mostly just style stuff#asahi-dev · 2026-08-02 10:02
svenyuka: does linux boot without the weird crashes with smp on 26.x or 27 now?#asahi-dev · 2026-08-02 10:01
chaos_princesssomething like that#asahi-dev · 2026-08-02 09:59
sofuslike use one tmp variable for each one?#asahi-dev · 2026-08-02 09:56
chaos_princessat least move the more complex logic expressions into variables#asahi-dev · 2026-08-02 09:55
sofusyeah, its bad... im not sure how to improve it#asahi-dev · 2026-08-02 09:52
chaos_princesschanging "wrong" to "borderline unreadable"#asahi-dev · 2026-08-02 09:51
chaos_princessok, nvm, i see.#asahi-dev · 2026-08-02 09:51
chaos_princesswhat is the intention there?#asahi-dev · 2026-08-02 09:50
chaos_princesshttps://github.com/AsahiLinux/linux/pull/561/changes/a5827b72c54b0e853cce205f83acc740294f6e1a#diff-c1c988a4f64dfa5773b9720c0a0851ed389a573d18cb297efedfcb64cb1e6d4fR543 - this part is probably wrong#asahi-dev · 2026-08-02 09:49
yukahttps://github.com/yuyuyureka/m1n1/commit/fd8832414172cf62b1396c73eacbf6ec08524192#diff-c4e733aec75a764a3a50d6181197db87e4c846891af23dec962b9592e96c858cR4-R352#asahi-dev · 2026-08-02 09:48
yukaI read somewhere that the nvme stuff changed starting from 15.x, so I was about to check if the changes I made for 26.6 also work on 15.x so that I can use the correct conditional#asahi-dev · 2026-08-02 09:47
yuka26.6 is good on M4* and A18 Pro. 27.x will probably be required for M5*#asahi-dev · 2026-08-02 09:46
chaos_princess15 only gets us the m3 ultra, and there are probably not that many of them, so it is likely will be bundled with m4+#asahi-dev · 2026-08-02 09:46
chadmedbut we know 26.6 works enough#asahi-dev · 2026-08-02 09:46
chadmedor 27 if they fix more broken sysregs #asahi-dev · 2026-08-02 09:46
yukaAh ok#asahi-dev · 2026-08-02 09:46
chadmed^ #asahi-dev · 2026-08-02 09:46
chadmedif we cant use 14.x we should probably just align on 26.x#asahi-dev · 2026-08-02 09:46
chaos_princess26.6#asahi-dev · 2026-08-02 09:45
yukaWhich 15.x version would we attempt to use for M3 Ultra?#asahi-dev · 2026-08-02 09:37
chaos_princessyes, that is for entropymode.c#asahi-dev · 2026-08-02 09:37
sofusi dont fully get it, but cool! Would it actually make sense to do this in the entropymode.c file? right now its easy to see the difference between our and verisilicon's#asahi-dev · 2026-08-02 09:36
chadmed(where they are repeated expansions of the same coefficients of course, doesnt help where the coefficients are unique) #asahi-dev · 2026-08-02 09:36
chadmedyou could probably also define all these repeating blocks of coefficients as macros where they need to be used so you dont need to repeat the repeated macros every single time#asahi-dev · 2026-08-02 09:32
chaos_princessuse this trick so you always use AOM_CDF at the usage site without needing to count arguments#asahi-dev · 2026-08-02 09:31
chaos_princesshttps://godbolt.org/z/nTTYa71M8 #asahi-dev · 2026-08-02 09:31
chadmedwe can probably get that down to like 9.8% with a bit more preprocessor geneva suggestions violations #asahi-dev · 2026-08-02 09:30
sofuswhen this gets merged avd is gonna be responsible for like 10% of our loc :p#asahi-dev · 2026-08-02 09:27
sofusmacos conformance for reference: https://pastebin.com/aC103V7H#asahi-dev · 2026-08-02 09:26
sofusthe long lines are very annoying, so also some better way to name stuff if someone has a suggestion for that#asahi-dev · 2026-08-02 09:25
svenhow could anyone say no to that ;)#asahi-dev · 2026-08-02 09:25
sofuschaos_princess: i tried to match the overall style of the other drivers, but sure#asahi-dev · 2026-08-02 09:24
chaos_princesssofus: do you want suggestions for more preprocessor abuse#asahi-dev · 2026-08-02 09:24
sofusactually not, most of that comes from verisilicon/rockchip_av1_entropymode.c#asahi-dev · 2026-08-02 09:24
sven*:D#asahi-dev · 2026-08-02 09:24
svenso kinda like atcphy is 80% weird magic #defines and register sequences :d#asahi-dev · 2026-08-02 09:23
chadmedits almost all static coefficient definitions :P #asahi-dev · 2026-08-02 09:22
sven6k lines, wow. that must be one of the larger drivers#asahi-dev · 2026-08-02 09:22
sofuslinux: https://pastebin.com/cZe7HsaL#asahi-dev · 2026-08-02 09:22
chadmednice work!#asahi-dev · 2026-08-02 09:22
svennice!#asahi-dev · 2026-08-02 09:21
sofusav1 still needs some love, but i would consider this "good" enough for now#asahi-dev · 2026-08-02 09:21
sofusim not very happy with the style changes, i hope someone has some input on that#asahi-dev · 2026-08-02 09:20
sofusav1 support https://github.com/AsahiLinux/linux/pull/561#asahi-dev · 2026-08-01 23:09
yukaT8132/T6040 pcie is super annoying. On T8132 the ethernet no longer wants to show up at all, while on T6040 it appears followed by "link down" and failed to change from D3hot to D0? I still didn't try if port 0 actually works for wifi. I now have a trace for T8132 too and identified some differences but adapting for those didn't help.