#asahi-dev · 2022-06-26 15:06
yunisven: ^_^#asahi-dev · 2022-06-25 19:56
yuniAll I know for certain is that M1 is G13G B0#asahi-dev · 2022-06-25 19:55
yuniDelightful#asahi-dev · 2022-06-25 17:58
yuniLooks like @woachk added those names#asahi-dev · 2022-06-25 17:57
yuniI'm pretty sure the H13J names on the wiki are wrong, I don't see where they came form#asahi-dev · 2022-06-25 17:55
yuninot super illumintating#asahi-dev · 2022-06-25 17:54
yunireveals that 13_3 is 13_3, A12 is both A12 *and* A12X, G13S_A0 is A14X, and G13X is also A14X#asahi-dev · 2022-06-25 17:54
yuni$ ls /System/Library/Extensions/AGXMetal*.bundle/Contents/Resources/*Statistics*derived.js | cat#asahi-dev · 2022-06-25 17:50
yuniThat would imply that the H13J names in the wiki are wrong#asahi-dev · 2022-06-25 17:50
yuniExcept some early stepping of one of them (the "A0" stepping of "G13S") had some hardware bug requiring a workaround in userspace, and the workaround was invasive enough to compile a separate variant of the driver?#asahi-dev · 2022-06-25 17:49
yuniAnd ordinarily all three share a userspace driver, AGXMetalG13X#asahi-dev · 2022-06-25 17:49
yuniAll three share a kernel driver, AGXG13X.kext#asahi-dev · 2022-06-25 17:49
yuniMaybe M1 Pro, M1 Max, and M1 Ultra are all G13X where X denotes "some letter other than P or G"#asahi-dev · 2022-06-25 17:48
yuni*G11G#asahi-dev · 2022-06-25 17:48
yuniFor the kernel driers, I see G131G, G13G, and G13X only.#asahi-dev · 2022-06-25 17:48
yuni...Wait. Those were the Metal drivers.#asahi-dev · 2022-06-25 17:47
yuniSo maybe Pro/Max are G13X and Ultra is G13S or something like that?#asahi-dev · 2022-06-25 17:47
yuniThat leaves A14 (H13P), M1 Pro (H13J), M1 Max (H13J), and H13J (M1 Ultra) within the 13 family#asahi-dev · 2022-06-25 17:46
yuni13_3 is H13G, I believe#asahi-dev · 2022-06-25 17:46
yuniA12 is clearly A12#asahi-dev · 2022-06-25 17:45
yuniComparing against https://github.com/AsahiLinux/docs/wiki/Codenames doesn't answer much#asahi-dev · 2022-06-25 17:40
yuniI'm particularly curious -- do we know what G13S_A0 is for?#asahi-dev · 2022-06-25 17:39
yuni13_3, 13_3_a, A12, G13S_A0, G13X#asahi-dev · 2022-06-25 17:39
yuniIn /System/Library/Extensions on 12.x, I see many AGX drivers:#asahi-dev · 2022-06-23 23:10
yunisven: woo!#asahi-dev · 2022-06-21 14:58
yuniYeah, that's fair#asahi-dev · 2022-06-21 14:43
yuniOh, I assumed M2 would be 13, alright#asahi-dev · 2022-06-21 14:32
yuniYeah, I just meant ABI changes for 12->13#asahi-dev · 2022-06-21 14:05
yuniwonder how bad DCP will be to uprev#asahi-dev · 2022-06-19 23:59
yuni"Move Apple DART", I mean#asahi-dev · 2022-06-19 23:59
yunijannau: Hopefully that also lets you handle the dcp darts in a less-gross way?#asahi-dev · 2022-06-19 20:23
yunisven: All the drivers are a mess ^_^#asahi-dev · 2022-06-19 14:43
yunisven: good luck! :-)#asahi-dev · 2022-05-23 13:03
yuniThe idea is also frequently used in page tables.#asahi-dev · 2022-05-23 13:03
yuniWhere the bottom bit distinguishes the Thumb encoding from the regular encoding.#asahi-dev · 2022-05-23 13:02
yuniAs Jamie[m] correctly jokes, this scheme is used for instruction pointers on certain Arm CPUs.#asahi-dev · 2022-05-23 13:01
yuniLike the type of data that's being pointed to, or the length (itself divided by a power-of-two, only storing the high bits), or so on.#asahi-dev · 2022-05-23 13:00
yuniThen, where you would otherwise have those zero bits, other pieces of data can be squeezed in.#asahi-dev · 2022-05-23 12:59
yuniDepending on the required alignment, that could be up to 6 bits that aren't stored and are assumed to be zero.#asahi-dev · 2022-05-23 12:59
yunimillenialhacker: Because addresses are aligned, it's common not to store the bottom bits.