#asahi-dev · 2023-01-04 15:58
ketteniscrazy#asahi-dev · 2023-01-03 13:12
kettenisthose are basically only used for the GPUs#asahi-dev · 2023-01-03 13:11
kettenisI had a quick look the other day, and the only iommu drivers that seem to do runtime pm are the SoC-specific ones#asahi-dev · 2023-01-03 13:10
kettenisjannau: which other iommu drivers?#asahi-dev · 2023-01-03 11:27
kettenisbut that does mean there is a dependency that needs to be enforced#asahi-dev · 2023-01-03 11:25
kettenisif DMA is quiesced before the iommu is powered down, it should work#asahi-dev · 2023-01-02 23:16
kettenisthat's kinda bad isn't it?#asahi-dev · 2023-01-02 11:25
kettenissame fundamental idea; a property on the connector#asahi-dev · 2023-01-02 11:25
kettenissame fundamental idea; a #asahi-dev · 2023-01-02 11:25
kettenisyes, as (fairly minimal) patches to the drm code#asahi-dev · 2023-01-02 11:16
kettenisah, interesting, I have drm integrated backlight handling on OpenBSD ;)#asahi-dev · 2023-01-02 10:57
kettenis(we just integrated drm from linux 6.1 into OpenBSD)#asahi-dev · 2023-01-02 10:55
kettenisis your current driver based on 6.1? or did you already move ahead?#asahi-dev · 2023-01-02 10:55
kettenisjannau: not looked at it in detail yet#asahi-dev · 2023-01-01 17:45
kettenissven: great, thanks#asahi-dev · 2023-01-01 16:43
kettenismarcan: note that my measurements are with the clusters already in their lowest P-state#asahi-dev · 2023-01-01 16:36
kettenisso that doesn't make me think this is a really deep idle state#asahi-dev · 2023-01-01 16:36
kettenisbut we only lose (most of) the GPRs in that state; not the system registers, TLBs and caches#asahi-dev · 2023-01-01 16:35
kettenisthat gives somewhere between 0.1 and 0.15 W extra power savings I'd say, but indeed I'd expect more if the whole cluster was powered down#asahi-dev · 2023-01-01 16:34
kettenismarcan: I'm fairly confident my secondary cores are in WFI with the "ok2pwrdn_force_down" bits set in SYS_IMP_APL_CYC_OVRD#asahi-dev · 2023-01-01 16:19
kettenisif you can't find someone who knows how this works in Linux, you might want to take a look at the gpio drivers for the SoCs used on $FRUIT pi boards#asahi-dev · 2023-01-01 16:16
kettenisexporting gpios to userland is certainly possible in Linux, but it may require a little bit of additional code to the relevant drivers#asahi-dev · 2023-01-01 16:11
kettenisbut beyond exporting gpios, I don't think linux has any infrastructure for this#asahi-dev · 2023-01-01 16:11
kettenisI can't think of a good reason why you'd want to do so#asahi-dev · 2023-01-01 13:20
ketteniswhen idle (now with the screen off)#asahi-dev · 2023-01-01 13:20
kettenisSYS 6643164, CPU 3820596#asahi-dev · 2023-01-01 13:20
ketteniswhen suspended#asahi-dev · 2023-01-01 13:20
kettenisSYS 4574661, CPU 2538621#asahi-dev · 2023-01-01 13:19
kettenisoh, actually those figures were measured with a usb cable connected causing the machine to think power is connected#asahi-dev · 2023-01-01 13:16
kettenisbut obviously at the point I call the SMC code the primary CPU (an e-core) is running, so no longer into that deep state#asahi-dev · 2023-01-01 13:16
kettenismy current OpenBSD hack puts all cores in this "deep WFI" state#asahi-dev · 2023-01-01 13:12
kettenisguess your 5W is from m1 max?#asahi-dev · 2023-01-01 13:11
kettenismachine idle, with the display on#asahi-dev · 2023-01-01 13:10
kettenisSYS 14909630, CPU 3311439#asahi-dev · 2023-01-01 13:10
kettenisbefore suspend this was#asahi-dev · 2023-01-01 13:10
kettenisSYS=PSTR, CPU=PHCP#asahi-dev · 2023-01-01 13:10
kettenisCPU 2501957#asahi-dev · 2023-01-01 13:10
kettenisSYS 5405363#asahi-dev · 2023-01-01 13:03
ketteniscan try this code on m1 pro#asahi-dev · 2023-01-01 13:03
kettenisPSTR reports ~1.9W#asahi-dev · 2023-01-01 13:02
kettenisPHCP reports something like 0.77W when I come out of "suspend" with one e-core running code#asahi-dev · 2023-01-01 13:01
kettenisthis is on M2#asahi-dev · 2023-01-01 12:46
kettenisbut yes, we really want to power down the cluster completely for suspend#asahi-dev · 2023-01-01 12:45
ketteniss/write/wrote/#asahi-dev · 2023-01-01 12:45
kettenis(I write code similar to what you wrote for m1n1 when you were experimenting with the turbo states)#asahi-dev · 2023-01-01 12:44
kettenis"deep WFI" is the state where it loses the GPRs but otherwise stays alive#asahi-dev · 2023-01-01 12:32
kettenisputting the p-Cluster in the "deep WFI" state also saves a bit more than 0.1 W#asahi-dev · 2022-12-19 12:33
kettenisright#asahi-dev · 2022-12-19 12:18
kettenisapalos: here now#asahi-dev · 2022-12-18 17:56
kettenisyour plan is to still follow through on that?