#
✕ nick:nickchan
#asahi-dev · 2025-02-12 02:07
nickchanokay macOS is safe because the MIDR list doesn't have m3
#asahi-dev · 2025-02-12 02:06
nickchanpossibly on macOS as well but not sure
#asahi-dev · 2025-02-12 02:06
nickchanif my understanding is correct then PMUv3 emulation won't go very well when combined with NV
#asahi-dev · 2025-02-11 11:25
nickchanstill not perfect however
#asahi-dev · 2025-02-11 10:12
nickchana9.plist and after seems to be okay however
#asahi-dev · 2025-02-11 10:12
nickchanon macOS /usr/share/kpep/a7.plist and /usr/share/kpep/a8.plist contains plenty of false information about pmu event affinities
#asahi-dev · 2025-02-09 12:27
nickchanlooks fine to me
#asahi-dev · 2025-02-09 11:52
nickchancause it's looking like an erratum now
#asahi-dev · 2025-02-09 11:52
nickchanmaybe that's fine to treat as all instructions?
#asahi-dev · 2025-02-09 11:50
nickchana7 also works but the event that supposedly counts all instructions turns out to only count a64 instructions
#asahi-dev · 2025-02-09 11:49
nickchanpmu works like a charm on a8-a10 now
#asahi-dev · 2025-02-09 11:49
nickchanwonderful
#asahi-dev · 2025-02-09 11:38
nickchanchaos_princess: AIC 1 and non-cpu0 boot core
#asahi-dev · 2025-02-09 11:37
nickchanunfortunately that feature is iphone only
#asahi-dev · 2025-02-09 11:37
nickchanand for a12x/a12z/t8027, its aic needs a mapping table since it boots on a core other than cpu 0
#asahi-dev · 2025-02-09 10:47
nickchanwell, other than poisoning the cache line
#asahi-dev · 2025-02-09 10:46
nickchanno way to run custom code on that in EL1
#asahi-dev · 2025-02-09 10:45
nickchanchadmed: 2019 ipad pro is a12x...
#asahi-dev · 2025-02-09 10:17
nickchanevery core has its own IRQ
#asahi-dev · 2025-02-09 10:17
nickchanokay I think I figured how AIC mode work now
#asahi-dev · 2025-02-09 10:15
nickchanFIQ is faster so I guess I should use FIQ
#asahi-dev · 2025-02-09 09:32
nickchana11 cpmu seems to be able to deliver IRQ and FIQ
#asahi-dev · 2025-02-09 09:20
nickchangot irq/aic mode to work for CPU PMU on a7
#asahi-dev · 2025-02-08 17:59
nickchanall of them are at least a month old without objections however
#asahi-dev · 2025-02-08 17:56
nickchansven: Everything outside of arch/arm64/boot/dts/apple (the bindings) have acks from krzk, as for the actual .dts files however, only the PMGR ones have some acks, excluding "arm64: dts: apple: t8012: Add PMGR nodes"
#asahi-dev · 2025-02-08 17:31
nickchansven: This is how I would resolve and apply the patches straight from the ML https://github.com/asdfugil/linux-apple/tree/ddbf3a964ef57b6da1149868235cf0574bcda2f6
#asahi-dev · 2025-02-08 17:30
nickchan3 and 4 conflicts with each other however because they add completely unrelated things to the same part of the .dts file (keep both is the resolution)
#asahi-dev · 2025-02-08 17:29
nickchan4. cpufreq nodes https://lore.kernel.org/asahi/20250203124747.41541-1-towinchenmi@gmail.com/T
#asahi-dev · 2025-02-08 17:29
nickchan3. PMGR DT bindings and nodes https://lore.kernel.org/asahi/20250203121831.36053-1-towinchenmi@gmail.com/T
#asahi-dev · 2025-02-08 17:28
nickchan2. T8012 device trees https://lore.kernel.org/asahi/20250203114417.16453-1-towinchenmi@gmail.com/T
#asahi-dev · 2025-02-08 17:27
nickchan1. split a9 variants device tree (1 patch) https://lore.kernel.org/asahi/20250203113949.14760-1-towinchenmi@gmail.com/T
#asahi-dev · 2025-02-08 17:27
nickchansven: 3 serieses and a singular patch, in order,
#asahi-dev · 2025-02-06 03:10
nickchan(...due to other issues like flat out not working properly on s5l8965x)
#asahi-dev · 2025-02-06 03:09
nickchanbut I need to do it anyways so maybe I should get to it
#asahi-dev · 2025-02-06 03:08
nickchanI hope it's still possible to integrate C code not in the mood of rewriting the dwc2 stack just yet
#asahi-dev · 2025-01-23 14:47
nickchanwhat is the values of AIDR_EL1 on m2, m3 and m4?
#asahi-dev · 2025-01-23 14:47
nickchanbit 12 /might/ represent GXF in AIDR_EL1 but very low confidence
#asahi-dev · 2025-01-23 14:43
nickchani don't think bit 16 in aidr_el1 represents gxf so the code currently in m1n1 is wrong
#asahi-dev · 2025-01-14 10:12
nickchanactually xnu uses aic mode on a7-a10
#asahi-dev · 2025-01-14 08:16
nickchanso it probably won't work
#asahi-dev · 2025-01-14 08:16
nickchanin the AIC mode that ostensibly is supposed to deliver an IRQ not only no IRQs are received but PMCR0_IACT never gets set as if it's in OFF mode
#asahi-dev · 2025-01-14 07:50
nickchanand then allow the custom sync exception handler to call the default sync exception handler
#asahi-dev · 2025-01-14 07:49
nickchanor maybe no hack but also give sync exception the fiq treatment when adding m1 support except the default handler is the normal sync exception handler
#asahi-dev · 2025-01-14 07:34
nickchanfor linux too
#asahi-dev · 2025-01-14 07:34
nickchanso it should be fine
#asahi-dev · 2025-01-14 07:34
nickchanthis is actually m1n1's behavior
#asahi-dev · 2025-01-14 07:27
nickchanthe FIQ handler probably should only touch the PMCR0 register when PMCR0_IMODE is set to FIQ
#asahi-dev · 2025-01-14 07:26
nickchanbecause I believe the CPMU has other modes even on M1
#asahi-dev · 2025-01-14 07:26
nickchanit's technically wrong for every CPU
#asahi-dev · 2025-01-14 07:26
nickchanactually

plus ›