#
✕ nick:chadmed
#asahi · 2022-05-19 03:28
chadmedwhich makes sense for the magsafe warning, though im not sure if iphone's magsafe is even strong enough
#asahi · 2022-05-19 03:27
chadmedyeah
#asahi · 2022-05-19 03:27
chadmedyeah the main risk from close contact is displacement
#asahi · 2022-05-19 03:25
chadmedin fact modern pacemakers wirelessly transfer data to a base station usually located next to the patient's bed. the magnetic flux that results from that data transfer is going to be far greater than anything the taptic engine could put out lmao
#asahi · 2022-05-19 03:23
chadmedlike these devices are designed to be 100% reliable in just about every possible environmental scenario for at least 10 years with no intervention or maintenance, because they are literally the only thing keeping some people alive
#asahi · 2022-05-19 03:21
chadmedyeah and even then im sure thats mostly just ass covering
#asahi · 2022-05-19 03:20
chadmeda good modern one? standing next to a really big electromagnet would probably screw with it a bit, or sticking rare earth magnets to your chest
#asahi · 2022-05-19 03:19
chadmedapple would simply never have been allowed to ship it if any deleterious physiological effects were observed as part of its use
#asahi · 2022-05-19 03:18
chadmedthe touchpad wreaking havoc on folks with pacemakers is just ludicrous
#asahi · 2022-05-19 03:18
chadmedyeah and i mean i can even link a study where a brain was subjected to an enormous 7T MRI field for over 24 hours continuously and no significant morphological changes were observed
#asahi · 2022-05-19 02:25
chadmedi dont know where you heard that a touchpad's haptic feedback motor is enough of an impulse to kick off perceptible physiological effects "particularly in people with cardiovascular disorders" but i suggest you stop getting information from there
#asahi · 2022-05-19 02:24
chadmed"it has physiological effects" theyre basically immeasurable at this level. very very few people even feel the effects of an MRI machine's magnetic field, which is anywhere from 0.5T to 3T in routine procedures
#asahi · 2022-05-19 02:23
chadmedktz_[m]: sorry i cannot let this slide. the idea that 30uT will affect ANYTHING let alone cardiac pacemakers is utterly preposterous
#asahi · 2022-05-14 04:38
chadmedtheir asahi installs will stay on 12.3.1 or whatever it is
#asahi · 2022-05-14 04:38
chadmedno
#asahi · 2022-05-12 01:33
chadmedbut very hard is a far cry from impossible
#asahi · 2022-05-12 01:33
chadmedfrom what stress testing i did before i had to get back to doing real life stuff, it was very hard to get the speakers in a state where they could theoretically let the magic smoke out
#asahi · 2022-05-12 01:32
chadmedhopefully we can start shipping speakers enabled in the dt by default soon
#asahi · 2022-05-09 11:32
chadmedno problem :)
#asahi · 2022-05-09 11:31
chadmedwork has been mostly focused on getting stuff landed upstream and tracing the GPU
#asahi · 2022-05-09 11:31
chadmedit is yeah
#asahi · 2022-05-07 02:46
chadmed"thermal pressure" is just a measure of the relative drop in performance caused by thermal throttling
#asahi · 2022-05-07 02:46
chadmedreading the LPC slide deck, it doesnt even sound useful outside of thermally throttled environments
#asahi · 2022-05-07 02:43
chadmedthis sounds very overengineered and cumbersome for SoCs that do not already rely on the kernel for performance statistics and power control
#asahi · 2022-05-07 02:36
chadmedif our goal is to optimise for performance and hardware block occupancy making the kernel chew through cycles deciding stuff like that is not a really good use of an AP's time
#asahi · 2022-05-07 02:36
chadmedthe more of this sort of logic you jam into the kernel the more dependencies you need to compile in (sensor drivers etc) which sort of brings up the lowest possible latency for any given system
#asahi · 2022-05-07 02:35
chadmedthat thermal stuff sounds like something that would be a good fit for a hardware scheduler tbh
#asahi · 2022-05-07 02:34
chadmedin terms of priority hinting for the scheduler, this is already implemented kind-of via niceness and i think there was some experimentation with explicit scheduler hinting added in code but idk what happened to that
#asahi · 2022-05-07 02:32
chadmedin high performance chips that dont need to as aggressively save power, you can sacrifice some of the power budget to a scheduler for an uplift in performance from more fine grained scheduling than the generic kernel mechanism can handle which improves occupancy
#asahi · 2022-05-07 02:31
chadmedfor high performance heterogeneous SoCs, vendors are probably going to start making greater use of hardware scheduling like Intel have done for Alder Lake
#asahi · 2022-05-07 02:29
chadmedall of which have to be maintained independently and you lose the collective benefit of scheduler improvements with how its done currently (generic mechanism everyone uses)
#asahi · 2022-05-07 02:29
chadmedif you wanted to optimise the software scheduler any further, you _could_ maybe rewrite the kernel system to just be a generic "give thread to x processor" mechanism that gets called by a vendor-specific scheduler driver but then you have approximately one morbillion schedulers for 3 squillion different SoCs
#asahi · 2022-05-07 02:26
chadmedthe problem is you cant really heavily optimise the scheduler for one system/set of cores or another because the kernel runs on such diverse hardware
#asahi · 2022-05-07 02:26
chadmedwe use a normalised performance scaling factor (nominally DMIPS/MHz but its just a dimensionless number representing the performance ratio between the core types)
#asahi · 2022-05-07 02:25
chadmedSobek[m]: i dont think theres much research left to do in terms of hetergeneous scheduling on linux. we know how to do it, and its probably about as good as software scheduling for these things can get
#asahi · 2022-05-03 01:36
chadmedthat lift VFD problem just sounds worse and worse the more i think about it :( reminds me of when i had to put up with DSL and my neighbour's cheapo solar inverter would knock out the internet every day between ~10:30am and ~2pm
#asahi · 2022-05-01 04:44
chadmed"virtual" gpu of sorts with its own render paths. there is nothing that the asahi driver can do to magically make software rasterisation faster, because it does not talk to llvmpipe
#asahi · 2022-05-01 04:44
chadmedllvmpipe is a
#asahi · 2022-05-01 04:44
chadmedthe kernel cant talk to the gpu yet, so the userspace part of the graphics stack is useless
#asahi · 2022-05-01 04:43
chadmedits not a magic handle into llvmpipe that is just faster on these machines
#asahi · 2022-05-01 04:42
chadmedthe asahi mesa driver isnt going to work until it has a kernel interface to the hardware
#asahi · 2022-04-23 02:34
chadmedoh easy change then :)
#asahi · 2022-04-23 02:07
chadmedshouldnt matter after u-boot has loaded the efi image
#asahi · 2022-04-23 02:07
chadmedm1n1 does pass some stuff via chosen but i think that's just to identify the ESP, shouldnt affect /boot
#asahi · 2022-04-23 02:06
chadmedit may be something hard coded in the removeable boot image
#asahi · 2022-04-23 02:04
chadmedyou could try that yeah though if youve definitely mounted your new boot partition at /boot that shouldnt matter
#asahi · 2022-04-23 02:01
chadmedand obviously ensure your kernels and initrds are all in /boot
#asahi · 2022-04-23 02:01
chadmedfurthermore after installing a new version of grub you need to run grub-mkconfig -o /boot/grub/grub.cfg
#asahi · 2022-04-23 02:01
chadmedit wont pick it up if you havent mounted it at /boot
#asahi · 2022-04-23 02:00
chadmeddid you then mount that to /boot in the filesystem and point the grub image to that?

plus ›