#asahi · 2025-07-31 12:40
chaos_princessand by one i mean at least (3 * cluster count) or so#asahi · 2025-07-31 12:40
chaos_princessthere are real energy counters in one of the pmgr mmios#asahi · 2025-07-31 12:37
chadmedthe SMC's power endpoints are a mix of derived and actual values. P**R are all measured as far as we know, and P**C are all derived#asahi · 2025-07-31 12:16
fl0_idI only tested with m3 and 14.x#asahi · 2025-07-31 12:16
fl0_idHowever it's likely that it would need to be adapted for m4 or newer macos versions#asahi · 2025-07-31 12:16
fl0_idhttps://github.com/step21/pymacjoule#asahi · 2025-07-31 12:16
fl0_idI repurposed this function calling code into a python module here: https://github.com/step21/pymacjoule/blob/main/src/lib.rs#asahi · 2025-07-31 12:14
fl0_iddruckerjunge you can look at what asimon and similar software does, which is calling private functions iirc. but the underlying mechanism is basically what chaos_princess said. And afaik it's not necessarily actual measurement, similar to RAPL, it's often an approximation. (afaik)#asahi · 2025-07-31 09:45
druckerjungeHmmm I somewhat expected this. But this definetly helps in understanding what I was seeing. Will take a deeper look into the documentation. Thank you, for you input.#asahi · 2025-07-31 09:44
chaos_princesshttps://developer.apple.com/documentation/iokit - this is how you access the data, or at least some of it#asahi · 2025-07-31 09:42
chaos_princessofc not, you'd need a kext to have direct access to a device and macos does not like new kexts#asahi · 2025-07-31 09:41
druckerjungeCurrently I'm not that deep into macos details. I thought about accessing some kind of device/register etc. using maybe elevated rights via a simple C++ interface.#asahi · 2025-07-31 09:36
chaos_princesshow do you intend to access them, from macos? if so, you can't, it is a hardware device that it wont give you direct access to. you will have to parse the ioreg database to get this data out#asahi · 2025-07-31 09:25
druckerjungeI came accross the "pmgr" device inside the CPU. The device shows several strings uppon dump like "total energy" or "ECPU0". So I guess there must be a way to read the actual values present in this component. I thought maybe one of you has already interacted with the energy meters on apple silicon and could share their experience. If this is not the right place, please excuse this long message. Kind regards#asahi · 2025-07-31 09:25
druckerjungeHey, maybe this might be a bit offtopic, as it is not directly related to asahi linux. I am hoping on getting information about interacting with the low level hardware of the mac book pro m4. Currently researching on the topic of energy sensors on apple silicon, similiar to Intel RAPL. I got my hands on an mac book pro M4. As far as I can tell, apple also integrated some kind of energy meter into the chip. By dumping ioreg #asahi · 2025-07-29 22:33
chaos_princessi think it also can sanely handle the situation when you have both fex and box64 installed, as binfmt_misc does not really handle multiple handlers for the same format that well#asahi · 2025-07-29 22:33
twb(Thanks for confirming)#asahi · 2025-07-29 22:32
twbRight, which binfmt-misc cannot express on its own. That's what I thought.#asahi · 2025-07-29 22:32
chaos_princesstwb: it is to skip muvm when host already has 4k pages#asahi · 2025-07-29 22:22
twbBut don't worry about it, I was just curious and it's not worth going aroud in circles#asahi · 2025-07-29 22:21
twbI was trying to ask what binfmt-dispatcher provide ON TOP OF the existing binfmt-misc which already handles scripts-without-shebangs and VM bytecode files#asahi · 2025-07-29 22:18
fl0_idtwb tbh I don't get what your remaining question is. Yes as far as I understand it's mainly for x86_64 dispatching. but minus universal binaries, like I have never seen those on linux even though it's prob possible#asahi · 2025-07-29 21:44
twbAt least on x86-64 Debian without binfmt-dispatcher, I have /proc/sys/fs/binfmt_misc/{cli,jar,python3.11,llvm-14-runtime.binfmt,qemu-aarch64,...}#asahi · 2025-07-29 21:42
twbHrm, perhaps that's only on Debian systems#asahi · 2025-07-29 21:41
twbRight but those are already implemented elsewhere#asahi · 2025-07-29 21:40
fl0_idthere are no fat / universal binaries afaik on linux. however if you actually read through some more, the underlying system is a general think, like f.e. also for pythoin and java code#asahi · 2025-07-29 21:31
twbOr is it also dealing with fat/universal binaries somehow#asahi · 2025-07-29 21:30
twbThat exists basically in order to insert muvm into the chain only when desirable?#asahi · 2025-07-29 21:28
fl0_idadriano here is in more detail what binfmt_dispatcher does https://github.com/AsahiLinux/binfmt-dispatcher#asahi · 2025-07-29 18:33
adrianomight honestly just build it myself#asahi · 2025-07-29 18:33
adrianono it is not, i did forget about that#asahi · 2025-07-29 18:33
chaos_princessis it an arm64 app?#asahi · 2025-07-29 18:32
adrianocedev-config now segfaults#asahi · 2025-07-29 18:32
chaos_princesstry uninstalling binfmt-dispatcher#asahi · 2025-07-29 18:31
adrianothe makefile takes the result of the binary cedev-config#asahi · 2025-07-29 18:31
adrianothe CEdev toolchain#asahi · 2025-07-29 18:30
chaos_princesswhat are you trying to run#asahi · 2025-07-29 18:29
adrianoim just wondering what binfmt_dispatcher does, since it seems to be causing me some issues with a program im trying to run#asahi · 2025-07-29 18:28
tpw_rulessure but it's a bit quiet#asahi · 2025-07-29 18:27
adrianohello! is this a good place for general questions about an issue? if not, are there any places like that?#asahi · 2025-07-29 11:42
fl0_idbtw, for backups, what I also do for other tools, but would work for time machine too - create a sparse disk image on the external hdd or nas or wherever and then you can use it for other usage, and still have the apfs or whatever stuff#asahi · 2025-07-28 23:45
twbOh like... time machine has a daemon that *serves* content on the old netatalk port? Weird.#asahi · 2025-07-28 23:45
linkmauveBut it did listen on port 548.#asahi · 2025-07-28 23:45
linkmauveOh.#asahi · 2025-07-28 23:44
twblinkmauve: you definitely don't need AFP for Time Machine anymore#asahi · 2025-07-28 23:39
linkmauveI tried running netatalk on my laptop but Time Machine didn’t accept its version: https://github.com/Netatalk/netatalk/issues/2304#asahi · 2025-07-28 23:38
linkmauveOr maybe their own implementation of it.#asahi · 2025-07-28 23:38
linkmauvetwb, it uses Ext4 and it apparently does AFP too, so probably running netatalk on Linux I’d guess.#asahi · 2025-07-28 23:37
linkmauvechaos_princess, off-the-shelf, from QNAP.