#asahi-re · 2026-06-18 07:45
JamesCalligeros[m]still has to bench m2#asahi-re · 2026-06-18 07:45
JamesCalligeros[m]Yeah they start pulling down unreasonable power at the higher pstates#asahi-re · 2026-06-18 07:43
JamesCalligeros[m]But im not the trillion dollar company cpu engineers i guess#asahi-re · 2026-06-18 07:43
JamesCalligeros[m]Point being i cant imagine theres much room between them in the power-perf curve to justify a whole third cluster#asahi-re · 2026-06-18 07:42
JamesCalligeros[m]Yes#asahi-re · 2026-06-18 07:39
JamesCalligeros[m]It’s a bit stupid. The ecores are already very close to the pcores performance wise#asahi-re · 2026-05-13 00:57
JamesCalligeros[m]either reimplement the functionality or document it and let someone else do it. we arent here to cut corners and attract apple's ire#asahi-re · 2026-05-13 00:56
JamesCalligeros[m]that notwithstanding, the APSL is just not compatible with the GPL so we cant just import code licenced under it into m1n1 wholesale#asahi-re · 2026-05-13 00:56
JamesCalligeros[m]"tooling" is throwaway/one-shot stuff like the AGX shader disassembler or the scipy thingy i use to trim zeroes out of FIRs. the hypervisor is not "dev tooling" in that sense and is a major component of m1n1, which is software that we release and distribute to users as part of the project#asahi-re · 2026-05-13 00:48
JamesCalligeros[m]we are happy for you to use Apple's code for experimentation and "tooling" but we cannot just integrate chunks of XNU into m1n1#asahi-re · 2026-05-13 00:48
JamesCalligeros[m]"but they cannot become an actual release of the project."#asahi-re · 2026-05-13 00:45
JamesCalligeros[m]apple's code is not permissively licenced. you need to reimplement the logic or document the logic and let someone else do it. #asahi-re · 2026-05-13 00:45
JamesCalligeros[m]no, that is never okay#asahi-re · 2026-05-10 06:55
JamesCalligeros[m]It’s SME v0.0.00001 pre release early access alpha :p#asahi-re · 2026-05-04 09:11
JamesCalligeros[m]theoretically posting snippets of decompiled code should be fine though, but yes please strictly limit it to this channel only#asahi-re · 2026-05-04 09:10
JamesCalligeros[m]screenshots are a bit annoying to deal with on the irc side#asahi-re · 2026-05-01 08:12
JamesCalligeros[m]If you squint really really really hard it kinda looks like smasnug’s codecs#asahi-re · 2026-04-02 06:53
JamesCalligeros[m]Just pick something that is TBA and get cracking i suppose. Have a look at the m1n1 codebase and documentation to see how we RE stuff#asahi-re · 2026-04-02 06:52
JamesCalligeros[m]That is it. I need to clean it up and updates but that is supposed to be the source of truth#asahi-re · 2026-04-02 06:43
JamesCalligeros[m]What are you interested in investigating?#asahi-re · 2026-03-29 02:57
JamesCalligeros[m]the colourmode being set is actually YCC422#asahi-re · 2026-03-29 02:57
JamesCalligeros[m]DCP_COLOR_FORMAT_YCBCR444 = 3 in parser.h. it is actually 2 in DCPAV.#asahi-re · 2026-03-29 02:56
JamesCalligeros[m]i may have found part of the problem #asahi-re · 2026-03-28 08:34
JamesCalligeros[m]aiui iomfb was one of the most attacked parts of xnu#asahi-re · 2026-03-28 08:34
JamesCalligeros[m]yeah thats what ive heard too, but they didnt have to move almost all of iomfb in to the firmware for that :P#asahi-re · 2026-03-28 06:42
JamesCalligeros[m]Though as preposterous as the firmware blob is, it is preferable to having to roll all this ourselves#asahi-re · 2026-03-28 06:41
JamesCalligeros[m]Easier to move it to the remote end of a dart than plug all the leaks in iomfb :p#asahi-re · 2026-03-28 06:41
JamesCalligeros[m]Its a lot more than that, its all the old kernel display out handling code rolled into firmware because it kept getting pwned#asahi-re · 2026-03-28 06:33
JamesCalligeros[m]I dont think the surface colourspaces and xfer function names we have are 100% accurate#asahi-re · 2026-03-28 06:32
JamesCalligeros[m]Ive stared at ghidra all day though so i think i will just pinch an hdmi adapter from work and trace dcpext to figure out ycc hdr#asahi-re · 2026-03-28 06:31
JamesCalligeros[m]I figured out that the byte below the fourcc is the range but there are a bunch of other bytes that look like bitfields#asahi-re · 2026-03-28 06:30
JamesCalligeros[m]Yeah#asahi-re · 2026-03-28 04:16
JamesCalligeros[m]the dcp firmware fourccs are actually part of a packed array of format descriptors#asahi-re · 2026-03-24 23:58
JamesCalligeros[m]yeah i mean at the end of the day, the question is always why not :P#asahi-re · 2026-03-24 23:52
JamesCalligeros[m]some stuff may work, some stuff may not, everything will be slow, and you will probably explode the kernel more than once#asahi-re · 2026-03-24 23:51
JamesCalligeros[m]its basically just undefined behaviour. the linux gpu drivers for the RGB mob are largely built assuming they can do write combining and reordering#asahi-re · 2026-03-24 23:43
JamesCalligeros[m]iirc the pcie controller can only map Device-nGnRE or Device-nGnRnE#asahi-re · 2026-03-12 04:29
JamesCalligeros[m]agx maybe?#asahi-re · 2026-02-19 23:01
JamesCalligeros[m]Oh i remember having a toshiba tecra laptop that could do that#asahi-re · 2026-02-10 06:28
JamesCalligeros[m]Sounds like its just CDC ACM given the “evidence” is /dev/ttyACM#asahi-re · 2026-01-13 01:33
JamesCalligeros[m]You need to use the raw m1n1.bin blob and configure it appropriately. Mach-O m1n1 broke some time ago and also wouldnt work as we would be under sptm anyway