#
…

2021-10-10

00:05doggkruse has joined #asahi
00:07hspak2 has quit [Quit: bye]
00:11hspak2 has joined #asahi
00:12timetheory[m]What’s the current usability status of Asahi Linux?
00:15hspak2 has quit []
00:18hspak2 has joined #asahi
00:29hspak2 has quit [Quit: bye]
00:42doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
00:47doggkruse has joined #asahi
01:14marcansven: OTOH it's rather unlikely that xarts will break with linux-apfs
01:14marcanit's a single purpose volume with like one file
01:14marcandoesn't get any simpler than that
01:14marcanso we can probably just ship it
01:15marcanI wouldn't tell people to run apfs as their rootfs, but we can probably do that much
01:17timetheory[m]Am I being ignored or is the matrix bridge broken?
01:18marcantimetheory[m]: I'm not sure what kind of answer you're looking for :)
01:18timetheory[m]marcan: I'll take that as being ignored ;)
01:19marcanI just released a detailed progress report mentioning what works and what doesn't the other day, does that help?
01:19timetheory[m]marcan: Hmm, I can't find it on the homepage or with a quick search, do you have a link handy by any chance?
01:21timetheory[m]Thank you so much!
01:23timetheory[m]When you state that "M1 Macs are actually usable as desktop Linux machines", does that mean that you'd actually recommend using Asahi as a daily driver or is it not stable enough yet?
01:23marcanit's not about stability, it's about features
01:24marcanI imagine most users would want e.g. audio to work
01:24timetheory[m]Whoops I missed that part
01:24timetheory[m]Though yeah the project has been coming along really quickly so I wish you guys best of luck
01:25timetheory[m]s/really/surprisingly/
01:25marcanI think we're doing a pretty good job at keeping things stable; I expect a bit of a longer tail there in graphics since that has a much larger bug surface, but otherwise things basically just work once they're implemented
01:26marcanmore likely is that people will run into corner cases that aren't properly supported
01:27marcanso all that will take some time to sort out :)
01:27timetheory[m]marcan: Honestly that's kinda the reason I want to use Linux instead of macOS; macOS seems to break when you do weird stuff but Linux seems pretty stable even when you break it
01:28timetheory[m]What are your thoughts about this continuing to work on next-gen macs? I saw some thoughts on reddit the other day that Apple may change courses and not allow custom OSs on their second generation like Sony
01:29marcanhighly unlikely to happen
01:29marcanreddit/HN/phoronix/etc are full of FUD
01:30marcanapple put together like a whole team of people to develop the boot policy system to make this happen
01:30marcanthey aren't just going to throw it all away
01:30timetheory[m]Wonder why they didn't release any docs at all :|
01:30marcanbecause the "official" use case is running custom XNU buids
01:31marcanwhich the XNU lead manager at Apple did blog about ;)
01:43chadmedi would also think that apple trying to lock down the bootloader of ostensibly desktop-class devices would open up a regulatory can of worms for them. it would be an incredibly significant break from the last 30-odd years of computer tradition and i cant see regulatory bodies like the EU taking too kindly to such locking down
01:44chadmedespecially considering the accelerating support for right to repair and device ownership among legislators and the like. maybe RMS will get his wish and tivoisation will eventually be outlawed
01:55doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:01robinp_ has joined #asahi
02:08robinp has quit [Ping timeout: 480 seconds]
02:09marcanFWIW, it's pretty clear that a lot of folks inside Apple are cheering for us
02:09marcancorporate is unlikely to make any official statements or guarantees, but they also have no reason to make things harder for us
02:11arahael1That's curious, and imho flies against the trends in Apple. However, that means it's very hopeful and promising that we might have a *fantastic* ARM-based laptop for us to use for linux. :)
02:11timetheory[m]marcan: I mean if they want to make Mac a closed system (like iOS, App Store revenues are huge) they absolutely would have a reason to make things harder
02:11marcanarahael1: it doesn't
02:11marcanpeople have a very twisted idea of large companies
02:11marcanI know that; I worked at Google for almost 3 years
02:12arahael1marcan: To be fair, my perspective is that as an iOS developer.
02:12arahael1 is now known as Arahael
02:12marcanbig companies aren't a homogeneous revenue-maximizing behemoth
02:12marcanthey are made of people, and people have their own opinions and ideas
02:13ArahaelI'm not disputing that. :)
02:13marcanyou can have bad stuff come from the top, but there's only so many things management can concern themselves with, if nothing else
02:13chadmedfeudal capitalism like people think exists in big tech isnt really a thing outside of east asia. sure, large corporations can be shitty and evil but most of this is enabled and even somewhat created by people _outside_ the company manifesting it. think of reddit's opinions on apple as nothing but a marketing factory. patently false as anyone who has spent any time with their more recent devices can tell you.
02:13i509vcb[m]marcan: It is definitely easy to assign some label of the fat cat but really all companies are made up of people who do have things they need to achieve in order to sustain themselves and their livelihoods
02:13i509vcb[m]This however does not justify bad decisions execs make to be fair.
02:13ArahaelIt's quite probable that the perspective you have from well, being part of this project, and being more involved with the _desktop_ OS is quite different to what I've seen as an iOS dev and dealing with the App Store.
02:14marcani509vcb[m]: it does not, but what I'm saying is that you can't assume the company is one homogeneous mechanism and you can extrapolate its behavior from highly visible decisions
02:14marcanArahael: I think they have a pretty strong line with the desktop OS
02:15chadmedi doubt tim cook is personally edge-case testing kmutil and going "hang on couldnt they use this to boot linux? stop this immediately!"
02:15marcanthey want people to be safe so they are making it *annoying* to bypass things (e.g. unsigned code)
02:15Arahaelmarcan: That's good to hear. :)
02:15marcanbut they aren't going to *lock it down* like iOS
02:15marcanthat's why our installer will never quite be as friction-free as I'd like it to, but I understand the reasons why
02:15marcanand I'm okay with that
02:15ArahaelThat's interesting. Honestly, sometimes I wish they did, would make IT teams in corporate less likely to install antivirus cruft.
02:16timetheory[m]I’m primarily worried that they’re looking to make an App Store for macOS though
02:16Arahaeltimetheory[m]: They already do.
02:16i509vcb[m]chadmed: to be fair there are people who have gotten linux to run on some very bizzare hardware
02:16marcanthey have
02:16marcanalso, consider the whole per-OS security state model of these machines
02:16timetheory[m]Sorry I meant App Store only
02:17marcanthat is *amazing*; I doubt corporate came up with that, it was probably driven from engineering, but I have no idea
02:17marcanbut it is a huge boon for users
02:17Arahaeltimetheory[m]: Strictly speaking, iOS isn't app-store-only. You can distribute and deploy your iOS apps on a website and have people install them that way. (It just costs more)
02:17Arahaelmarcan: What do you mean by that, there? "the whole per-OS security state model".
02:17marcanmacs are the only platform where you can have full freedom to run your own OS *and* the ability to run a locked down DRM system, with a simple reboot
02:17marcanthe security state is per-OS
02:17timetheory[m]Arahael: How? PWAs on iOS are not exactly implemented in a good way
02:18marcanso you can have a full secure macOS install, capable of running iOS apps and watching Netflix in 4K
02:18Arahaeltimetheory[m]: I'm not talking about PWA's.
02:18marcanand fully open Linux
02:18marcanand dual boot them
02:18marcanthat doesn't exist anywhere else
02:18Arahaelmarcan: That's very cool!
02:18marcanall other platforms use some sort of global switch, e.g. Android with the OEM unlock
02:18chadmedi509vcb[m]: true but my point is more that i doubt corporate gives a damn about the technical minutiae of the device, especially if its stuff like this that engineering can sneak through as necessary for development. as marcan has said in the past, its not an accident that the macs behave the way they do.
02:18timetheory[m]Arahael: What are you talking about then?
02:18Arahaeltimetheory[m]: Your regular boring old IPA's.
02:19chadmedits as if theres a trail of breadcrumbs left by the engineers for us to follow. im sure they themselves would love to have linux on these devices too
02:19timetheory[m]Arahael: You can’t run those though without hacks
02:19Arahaeltimetheory[m]: Yes, you can. I do.
02:20timetheory[m]How? You either need to sign them via AltStore (which is kind of a hack) or have a jailbroken phone, right?
02:20marcanchadmed: they technically kind of do; Apple uses Linux for silicon validation
02:20timetheory[m]timetheory[m]: W/o a paid account you can only install 3 AltStore apps anyways
02:20marcanprobably ugly corellium-style ports though, not upstreamable
02:20Arahaeltimetheory[m]: Nope. You can also get an Enterprise developer account, with an Enterprise certificate and use enterprise distribution methods. It's generally intended for internal apps for the enterprise.
02:21Namidairoquite a few manufacturers are jerks about android oem unlock, even if it is provided
02:21brentr123[m]Gotta love fastboot oem unlock
02:21chadmedyeah you absolutely can sideload apps on ios as long as you have the certificate for it. company i used to work for used some ghoulish bespoke job ticket app that i had to install and maintain on my own device. awful stuff.
02:21brentr123[m]What a wonderful command that is
02:22timetheory[m]Arahael: Having to pay $300 a year extra on top of the phone doesn’t exactly count as open
02:22Arahaeltimetheory[m]: I don't think we were ever discussing being able to do so for free.
02:22marcanbrentr123[m]: if only it didn't break SafetyNet :p
02:22Namidairoranging from not having it at all, making you sign requests on a web page, imposing 2 week delays, degrading devices by wiping out calibration partitions (bye bye camera), etc.
02:23timetheory[m]That’s often an entire half of the phone’s price and it’s also not possible to distribute to end users
02:23chadmedi had an HTC 10. shame that HTC were such arseholes about unlocking it because it was truly a fantastic phone, best integrated DAC in a smartphone at that time. but even after unlocking it there were sections of the bootloader that were totally locked down. im guessing this was to protect some rootkit/firmware killswitch they had
02:23marcanNamidairo: that's another thing about the macs, they're unbrickable
02:23brentr123[m]marcan: Ever heard of magiskhide?
02:23brentr123[m]It bypasses safetynet
02:23timetheory[m]marcan: Why though?
02:23marcanbrentr123[m]: there is no way to bypass safetynet on modern devices with TZ-tied signing that knows about unlock state
02:23brentr123[m]Also isn’t safetynet only flagged with root? Not with boot loader unlock
02:24marcantimetheory[m]: why what?
02:24marcanbrentr123[m]: there are two levels
02:24Namidairothey flag certain ro. variables such as bootloader unlock now
02:24marcanroot will get you full untrusted, oem unlock on modern devices will get you CTS fail
02:25timetheory[m]marcan: Forgot that IRC doesn’t show replies, was replying to the statement of Apple using Linux for silicon validation
02:25marcanmy LineageOS Nexus 4a passes basicIntegrity but not CTS for this reason
02:25marcanwhich is good enough for some apps that use SafetyNet, but e.g. not Google Pay
02:25Namidairothey're slowly moving to attestation in trusted compute
02:25marcantimetheory[m]: because Linux is easy-ish to port and makes for a very good hardware test case
02:25marcansame reason corellium did their port...
02:26timetheory[m]Don’t they already have XNU though?
02:26marcanlinux is easier to port than XNU and exercises things XNU does not
02:26marcanplus it's good to have two OSes to test with
02:27marcanI expect they especially use Linux for CPU silicon validation
02:27marcanso they probably don't have drivers for everything
02:27marcanbut they will run a ton of CPU stress tests and such
02:27PhilippvK has joined #asahi
02:27marcanit's also a good test of ARM spec compat (*cough* and they already broke that with E2H=1 *cough*)
02:28timetheory[m]marcan: do they really care if it’s compatible with the spec though since it’s not like they’re officially intending for anyone to be writing bare metal code
02:29marcanthey definitely care that it's compatible with the spec for userspace code
02:29timetheory[m]And stuff has to be ported to macOS anyways for user space
02:29marcanand ARM will get very angry if they do something horribly wrong for kernel mode
02:29marcanno, think: VMs
02:29marcanand there is lots of ARM64 userspace code you would *not* want to have to port to macOS
02:29marcan(e.g. sharing code with Android)
02:29chadmedwell it is ostensibly a condition of their architecture license that any core design they bill as "arm" has to be compatible. dunno why they were allowed to break spec in this case but im guessing theyve likely been warned internally not to do it again.
02:30timetheory[m]chadmed: They haven’t really been advertising it as ARM though, they’ve been calling it Apple Silicon
02:30Arahaelchadmed: And I don't think we know the specifics of their license, do we? They were technically one of the original founders, weren't they?
02:30marcanapple are definitely pushing things with impdef extensions not allowed by the spec, but as far as I know the only actual compat break so far is the E2H thing
02:31phiologe has quit [Ping timeout: 480 seconds]
02:31chadmedtimetheory[m]: youre thinking too small as well. its "fine" for desktop macos but apple has big plans for their silicon, namely HEDT and server (the latter being rumours but seems logical). if they want to be taken seriously in these spaces they cant be breaking arm spec
02:31marcantimetheory[m]: it's not about the branding; to be able to implement the ARM instruction set they need a license
02:31marcanthis isn't like USB where the only thing that matters is the logo
02:32marcanbut anyway, user/kernel mode both need to be compatible for VMs to work
02:32ArahaelIn any case, I'm excited for the future in computer hardware, oddly enough. I felt we had stagnated with the Intel monopoly of all this.
02:32marcanand they absolutely want to support running windows/linux in VMs
02:32chadmedalso what marcan just said. billing something as "arm" isnt about the marketing, its about having the right to implement the instruction set. it is not openly licensed, and apple wouldve signed a massive massive contract to acquire their perpetual architecture license that says arm can revoke it if apple break the spec.
02:33marcan(note that the E2H break is about hypervisor mode, so it doesn't apply to VMs, which is why they work)
02:33timetheory[m]I still never understood why archs can be copyrighted
02:33marcanthey can't
02:33chadmedthis is similar to how SPARC is/used to be licensed. you couldnt legally produce a "SPARC" core unless it passed validation tests and conformed to the ISA as set out in the manual. if your core broke this spec the SPARC Consortium (basically just Oracle) would not validate your core
02:33marcanthey can be patented though
02:34timetheory[m]ARM is patented?
02:34marcanall commercial ISAs are
02:35timetheory[m]Would that mean that 64 bit x86 is expiring in 2 years?
02:35timetheory[m]Since according to Wikipedia it was introduced in 2003
02:35marcanthe way this works is they keep adding extensions, so you can't release a modern x86 CPU since people expect things like AVX to work
02:35marcanbut yes, you could clone the original opteron if you wanted to in a couple years
02:35marcanwith SSE2
02:36marcansimilarly, I'm pretty sure ARM can't stop you from releasing an ARM7 compatible CPU or something, though they might try
02:37chadmedwatch them become patent trolls if the nvidia deal goes through :P
02:37marcanI sure hope that gets blocked
02:38ArahaelWhat's that nvidea deal?
02:38chadmedlikewise. thankfully that looks very likely. the UK isn't keen on it and nor is the EU
02:38marcanthey're trying to buy ARM
02:38chadmedArahael: nvidia is looking to buy Arm from SoftBank
02:38chadmedtheyve reached an agreement to do so for ~$40bn US, but the UK, EU, US and Chinese governments are in the process of putting a stop to it
02:38ArahaelA shame how nvidia has turned out. I remember more than 15 years ago I actually liked Nvidia.
02:38Namidairomeanwhile also implementing their gpu coprocessors in risc-v
02:39chadmednvidia was never a good company and they never made particularly good products.
02:39chadmedleather jacket man is easily one of the most disingenous and awful people in the industry
02:39i509vcb[m]at least amd caught up in raster performance this gen, although good luck getting a gpu
02:40ArahaelWell, 15 years ago I was a kid, really, but they seemed to have better drivers in those days.
02:40chadmedand just this week there have been grumblings from AIBs that theyre trying to resurrect the GeForce Partner Program to lock AIBs out of making good intel cards
02:45i509vcb[m]intel trying to enter the high end gpu space is definitely interesting to watch
02:46chadmedim interested only insofar as how much they will focus on improving their linux drivers and if their flooding of the market will shatter the gpu scalper bubble
02:56chadmed has quit [Remote host closed the connection]
03:05phireYou can actually get supprisingly far without AVX
03:05phireit's only in the last few years that a few games have started requiring it by default
03:09sorearphire: avx is disabled by fuse in Pentium-branded parts for market segmentation reasons -> software can't rely on it -> baseline ISA patents expiring soon ... huge own goal by itnel if i've understood it correctly
03:09nepeat has quit [Ping timeout: 480 seconds]
03:10phirein our testing, a few games end up accentally depending on it
03:11phirebut in general most applcations/games keep the SSE4 codepath around, if not an SSE2 codepath
03:30marvin24_ has joined #asahi
03:33marvin24 has quit [Ping timeout: 480 seconds]
03:36nepeat has joined #asahi
04:10tomtastic_ has quit [Quit: ZNC - https://znc.in]
04:10tomtastic has joined #asahi
04:31chadmed has joined #asahi
04:51doggkruse has joined #asahi
04:53steevalyssa: does the mali t860 only have opengl es support? no full opengl?
05:25myon98 has joined #asahi
05:35kode54Bonus: Mac software can't rely on avx either, especially as rosetta 2 doesn't implement it
05:37kode54it does, however, implement an incredibly slow 80 bit softfpu in case you're stupid enough to use x87 opcodes
06:01chadmedkode54: reminds me of the matlab x87 controversy a couple of years ago lmao
06:17hspak has quit [Quit: The Lounge - https://thelounge.chat]
06:17hspak has joined #asahi
06:27tomtastic has quit [Quit: ZNC - https://znc.in]
06:27tomtastic has joined #asahi
07:09amw has quit [Quit: WeeChat 2.3]
07:58ihaveamac has quit [Quit: fail]
08:00amw has joined #asahi
08:00ihaveamac has joined #asahi
08:32doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
08:45aleasto has joined #asahi
08:45aleasto has quit []
08:46aleasto has joined #asahi
08:48aleasto has quit [Remote host closed the connection]
08:48aleasto has joined #asahi
08:49aleasto has quit [Remote host closed the connection]
08:51aleasto has joined #asahi
09:12darkapex2 has joined #asahi
09:13darkapex1 has quit [Ping timeout: 480 seconds]
09:21sorear has quit [Ping timeout: 480 seconds]
09:21jabashque has quit [Ping timeout: 480 seconds]
09:21NekomimiScience has quit [Ping timeout: 480 seconds]
09:21tom-w has quit [Ping timeout: 480 seconds]
09:21enomem has quit [Ping timeout: 480 seconds]
09:21guan has quit [Ping timeout: 480 seconds]
09:21aeroraptor has quit [Ping timeout: 480 seconds]
09:21ZLSA has quit [Ping timeout: 480 seconds]
09:21brinly has quit [Ping timeout: 480 seconds]
09:21tom-w has joined #asahi
09:21stblassitude_ has joined #asahi
09:22Chainsaw_ has joined #asahi
09:22austriancoder_ has joined #asahi
09:22kendfinger has quit [Ping timeout: 480 seconds]
09:22cptcobalt has quit [Ping timeout: 480 seconds]
09:22Chainsaw has quit [Ping timeout: 480 seconds]
09:22Chainsaw_ is now known as Chainsaw
09:23ovf has quit [Ping timeout: 480 seconds]
09:23austriancoder has quit [Ping timeout: 480 seconds]
09:23kendfinger has joined #asahi
09:23cptcobalt has joined #asahi
09:23austriancoder_ has quit []
09:24philpax has quit [Ping timeout: 480 seconds]
09:24Ariadne has quit [Ping timeout: 480 seconds]
09:24stblassitude has quit [Ping timeout: 480 seconds]
09:24stblassitude_ is now known as stblassitude
09:25austriancoder has joined #asahi
09:25philpax has joined #asahi
09:28Ariadne has joined #asahi
09:29eichin has quit [Remote host closed the connection]
09:29nkaretnikov has quit [Remote host closed the connection]
09:29cptcobalt has quit [Remote host closed the connection]
09:29eric_engestrom_ has quit [Remote host closed the connection]
09:30steev has quit [Read error: Connection reset by peer]
09:30Vaughn has quit [Remote host closed the connection]
09:30jkkm has quit [Remote host closed the connection]
09:32jkkm has joined #asahi
09:33ZLSA has joined #asahi
09:33cptcobalt has joined #asahi
09:33eichin has joined #asahi
09:33brinly has joined #asahi
09:33sorear has joined #asahi
09:33nkaretnikov has joined #asahi
09:33eric_engestrom_ has joined #asahi
09:33steev has joined #asahi
09:33guan has joined #asahi
09:33NekomimiScience has joined #asahi
09:33jabashque has joined #asahi
09:33Vaughn has joined #asahi
09:33enomem has joined #asahi
09:33aeroraptor has joined #asahi
09:42ovf has joined #asahi
10:04jacoxon has joined #asahi
10:13jacoxon has quit []
10:13jacoxon has joined #asahi
10:17j_eyI'm not sure if this regmap stuff will end up being a bit of a hack. I think we only want to cache if it's an output register
10:35ZLSA_ has joined #asahi
10:35nsklaus_ has quit [Quit: WeeChat 3.2]
10:35jabashque has quit [Ping timeout: 480 seconds]
10:35ovf has quit [Ping timeout: 480 seconds]
10:36eichin has quit [Ping timeout: 480 seconds]
10:36sorear has quit [Ping timeout: 480 seconds]
10:37steev has quit [Ping timeout: 480 seconds]
10:38ZLSA has quit [Ping timeout: 480 seconds]
10:38ZLSA_ is now known as ZLSA
10:39cptcobalt has quit [Ping timeout: 480 seconds]
10:39NekomimiScience has quit [Ping timeout: 480 seconds]
10:39brinly has quit [Ping timeout: 480 seconds]
10:39enomem has quit [Ping timeout: 480 seconds]
10:39jkkm has quit [Ping timeout: 480 seconds]
10:39guan has quit [Ping timeout: 480 seconds]
10:39jkkm_ has joined #asahi
10:39Vaughn has quit [Ping timeout: 480 seconds]
10:39aeroraptor has quit [Ping timeout: 480 seconds]
10:39eric_engestrom_ has quit [Ping timeout: 480 seconds]
10:39nkaretnikov has quit [Ping timeout: 480 seconds]
10:39brinly has joined #asahi
10:41NekomimiScience has joined #asahi
10:41aeroraptor has joined #asahi
10:49ovf has joined #asahi
10:49eichin has joined #asahi
10:50sorear has joined #asahi
10:51jabashque has joined #asahi
10:55steev has joined #asahi
10:55cptcobalt has joined #asahi
10:55Vaughn has joined #asahi
10:56eric_engestrom_ has joined #asahi
10:56nkaretnikov has joined #asahi
10:58guan has joined #asahi
10:58enomem has joined #asahi
11:06nsklaus has joined #asahi
11:18gabuscus has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
11:19gabuscus has joined #asahi
11:21jacoxon has quit []
11:44turbotum has quit []
12:08chadmed has quit [Remote host closed the connection]
12:09chadmed has joined #asahi
12:14BlitzWorks has quit [Quit: Gateway shutdown]
12:24BlitzWorks has joined #asahi
12:56ServerStatsDiscoverertraveler4 has joined #asahi
13:00marcanj_ey: in general caching is not bad, avoids reads
13:00marcan(which are slower since they stall)
13:00marcanthough this fabric is nGnRnE so writes stall too, but not as badly since they have no deps
13:02j_eymarcan: yeah, but I need to bypass the cache when the register is an input
13:02marcanhopefully regmap can do that for you, otherwise it'd be quite broken
13:04j_eyI'm trying to figure out the best way of doing that, there doesn't seem to be a regmap_read_no_cache function.
13:05j_eyI can (so far that I've found), either turn off caching for the entire map (so maybe need some locking), or using a callback function I can say if a register is volatile or not (and only say its volatile if its an input reg)
13:06j_ey(those are regcache_cache_bypass or the volatile_reg callback)
13:09alyssasteev: Mali GPUs can implement full opengl with varying amounts of hacks. Arm does not do so. Panfrost does a partial "as-is" OpenGL 3.1 implementation (not conformant but good enough for common linux software)
13:15marcanj_ey: if it doesn't implement caching on demand only then maybe it's not a good fit
13:16j_eyI will talk to the regmap maintainer, if I can't work it out soon
13:17marcanyou can just reply to the review email with your findings
13:19j_ey(regmap maintainer works at the same place as me, so hopefully I can get help :P)
13:36jvoisin has quit [Remote host closed the connection]
13:51j_eyoh.. I just realised I can do a normal readl in the non-cache case, and dont use regmap for that one call
13:59alyssaj_ey: but then the cache will be stale?
14:01j_eyhm. I clearly need to read more about the regmap APIs!
14:06j_eyother GPIOs hw are different in that the registers are split differently. in apple's pinctrl a register has all its value in one place. other GPIO/pinctrl have registers like: GPIO_DATA where each bit corresponds to the output data of pin n
14:06j_eyso you can say GPIO_DATA is a 'volatile' register in regmap
14:12___nick___ has joined #asahi
14:25marcanalyssa: but the cache doesn't care about the read bit
14:25marcanthe cache is for writes
14:25marcanyou don't *want* to update it when you read the pin
14:29j_eymarcan: here's what I have right now (untested) https://paste.gg/p/anonymous/8ce383b0d2b84f82a99b7fcd466f4b1d
14:30marcanthat looks weird
14:30marcanis that function supposed to return the written value for outputs?
14:30marcanor is it only used for inputs?
14:31marcanif it's supposed to work for outputs then I guess that makes sense
14:31j_eyuhh, good point
14:31j_eyI need to look into that too, I think it might be for both
14:42marcanyay, I think I fixed the i2c/hpm init fail in m1n1
14:43j_ey(at least corelliums keyboard code seems to call get() on an output reg, but that might just be wrong, will look at more examples)
14:55kov has joined #asahi
15:15povik has joined #asahi
15:25jacoxon has joined #asahi
15:29chadmed has quit [Remote host closed the connection]
15:29chadmed has joined #asahi
15:40jacoxon has quit []
15:49povik has quit [Remote host closed the connection]
16:19steevalyssa: ah, crud, i was hoping for 3.3 like freedreno so i could use alacritty :)
16:22psykoseif you want something similarly minimal you can try foot
16:23steevyeah, i'm using foot, i just prefer alacritty
16:42j_eyalyssa: can you test the new pinctrl patch (when its ready), before I send it out?
16:55tomtastic has quit [Ping timeout: 480 seconds]
16:58tomtastic has joined #asahi
17:10___nick___ has quit []
17:21bytesize0[m] has joined #asahi
17:25povik has joined #asahi
17:26povikhm, the oftc webchat client disconnected me silently
17:26povikanyway, i pulled pmgr changes and bought a better flash drive for rootfs
17:26povikthis is... pretty fast
17:29poviki am not running anything graphical
17:29povikbet the difference would be more stark if i did
17:30tomtastic has quit [Ping timeout: 480 seconds]
17:31tomtastic has joined #asahi
17:33povikalthough the primordial sound card disappeared, and i dont know why
17:33povikso so far i see this as net loss
17:54marcanpovik: I2S probably depends on the SIO power domain, you'd better add that power-domain especially if you don't have console on UART (which would also keep it alive)
17:54povikalready did
17:54povikthis is something that has to do with linux more than with hardware
17:54povikthe sound card does not register with alsa
17:55marcanjust caused by the linux version then, probably?
17:55poviki did change the base linux version if that's what you mean
18:04tomtastic has quit [Ping timeout: 480 seconds]
18:06tomtastic has joined #asahi
18:06jkkm_ is now known as jkkm
18:23alyssamarcan: Mm, ok.
18:23povik has quit [Remote host closed the connection]
18:23alyssasteev: The only big feature missing for GL3.3 is geometry shaders which are unlikely to ever be supported... I don't /think/ alacritty uses them
18:24alyssaIn theory you could set env var `PAN_MESA_DEBUG=gl3` and get GL3.3 that'll crash if you try to use geometry shaders or whatever
18:25alyssaIn practice that'll work for alacritty but break some other things so don't tell me if you do it, only do it on an app-specific basis (there is never a good reason to put PAN_MESA_DEBUG in a system-wide environment), and don't complain if it breaks ;-)
18:27alyssaj_ey: Not this weekend (it's Thanksgiving), maybe Tuesday
18:28j_eyalyssa: ok! have a fun thxgiving
18:33j_eythere's one bit in particular that I currently can't test
18:34dottedmagalyssa: Does macos also skip this bit of OpenGL in their macos->metal translation layer?
18:34dottedmag*opengl->metal
18:52alyssaj_ey: which bit?
18:54j_eywell all the PCIe stuff I dont have, but I meant there's one particular change that I would like to try
18:56alyssaah, sure
18:56alyssatechnically you should have wifi on airs with lots of patches, and that would exercise pcie, but..
18:57j_eyexactly :D
19:00svenethernet should just work as well
19:00svenoh wait
19:00svennevermind
19:00svenj_ey: i can test ethernet and usb a if you give me a patch tomorrow
19:10j_eysven: ty
19:19manjaro-mate has joined #asahi
19:20manjaro-mate has left #asahi [#asahi]
19:21joske has joined #asahi
20:16steevalyssa: nope, i'm posting on twitter that you said to always set PAN_MESA_DEBUG=gl3
20:19joske has quit [Ping timeout: 480 seconds]
20:20linearcannon has quit [Ping timeout: 480 seconds]
20:25jn has quit [Quit: No Ping reply in 180 seconds.]
20:25steevalyssa: seriously though, thanks. i had to do that with freedreno (or a similar) before we got full 3.3, but you're right, alacritty doesn't, it just claims to need it
20:26jn has joined #asahi
20:26jacoxon has joined #asahi
20:31psykosethere's a patch near the bottom that lets you drop the debug var.. but it's gonna be the same
20:32psykoseindeed the features that alacritty needs past 3.1 are supported in panfrost
20:39kettenisa terminal emulator that requires, OpenGL, seriously?
20:40kettenisI suppose if all you have is a hammer...
20:40psykoseopengl makes more sense in a terminal than the entire rest of the terminal stack
20:41i509vcb[m]Not sure if alacritty needing 3.3+ is a making of alacrity or glium
20:42psykosealacritty
20:43steevthere used to be a fork that had gles2 and gles3, but sadly, that person deleted their github, and the archive stuff doesn't build correctly
20:46linearcannon has joined #asahi
20:48ahmad has joined #asahi
20:55alyssasteev: enjoy your overengineered terminal gpu stack :)
20:56steevbut it's written in rust so it's okay :P
20:56psykoseoverengineered feels like the wrong word
21:04phireWell, 2d acceleration hardware doesn't exist anymore. If you want a fast terminal, you need to use 3d acceleration
21:05psykosei can't say i notice any speed difference whatsoever between foot and alacritty
21:05psykosebut it's certainly one way to get there
21:07steevmy big thing is, i already have an alacritty config i use everywhere else, i don't feel like customizing foot like that
21:07kettenisa modern CPU is plenty fast to render the contents of your terminal
21:07kettenisthe only thing you really need to do the GPU for is compositing
21:08ahmad has quit [Remote host closed the connection]
21:08psykoseyou say this but you have not felt the touch of windows terminal
21:08steevor havent used some arm processors
21:09phireThe increased preformance of alacritty/kityy is noticable when scrolling in vim
21:09phireAnd can actually increase the execution speed of programs that dump a lot of output to stdout
21:12phireThough, IMO the vim (and other editors) scrolling speed should be solved by creating modern escape codes that tell the terminal to move whole blocks of the screen.
21:16jacoxon has quit []
21:16jacoxon has joined #asahi
21:44tomtastic has quit [Quit: ZNC - https://znc.in]
21:45tomtastic has joined #asahi
22:00yrlf has quit [Quit: The Lounge - https://thelounge.chat]
22:01yrlf has joined #asahi
22:11jacoxon has quit []
22:27klangephire: That _is_ how it works.
22:46l3k[m] has joined #asahi
22:54l3k[m] has left #asahi [#asahi]
23:08aleasto has quit [Quit: Konversation terminated!]
23:49alyssaklange: in PenguinOS or in PonyOS?
23:51klangeIn everything. ^[[S and ^[[T are literally the escape sequences for moving whole blocks of the screen and that's how vim and everything else scrolls.
23:51alyssaspeaking of PonyOS for M1 when
23:51klangeI have an RPi4 finally arriving today which will get the basics of aarch64 support going, so "eventually and not as a joke".
23:53alyssahow about sortix, is that still kicking?
23:53klangesortie is quite actively working on sortix, lots of exciting stuff happening there
23:53alyssanice :-)
23:54klange(also ^[[L and ^[[M are for inserting and deleting lines, which may also be how it works, depending on terminfo)
…