channel logs for 2004 - 2010 are archived at http://tunes.org/~nef/logs/old/ ·· can't be searched
#osdev2 = #osdev @ Libera from 23may2021 to present
#osdev @ OPN/FreeNode from 3apr2001 to 23may2021
all other channels are on OPN/FreeNode from 2004 to present
http://bespin.org/~qz/search/?view=1&c=osdev2&y=24&m=4&d=22
00:16:00 <vaihome-> geist me? you mean the machine?
00:34:00 <vaihome-> vaihome- the shop opens at 10 AM, so I get money at around 13-14 afternoon
00:34:00 <vaihome-> thats when I am going to getit
00:39:00 <Ermine> Actually what it takes to add kmsan to an os?
00:41:00 <heat> the code
00:43:00 <heat> kmsan and kasan (and ktsan but *not* KCSAN) are basically dependent on shadow memory to store metadata onto
00:43:00 <heat> so if you have the code implemented (basically a bunch of callbacks), it's done as long as the compiler codegens properly
00:44:00 <heat> not that the sanitizers are super trivial, but it's not amazingly hard either
00:54:00 <Ermine> From kernel docs it seems like those callbacks are like __msan_instrument_asm_store ?
00:56:00 <heat> yep
00:56:00 <heat> for asan __asan, for ktsan __tsan, for ubsan __ubsan
00:56:00 <heat> you get the drill :)
00:58:00 * Ermine writes that to backlog of things to do after diploma
04:53:00 <klys_> just bought a pair of dec rx02 drives today.
04:53:00 <geist> no way raelly?
04:53:00 <geist> whatcha gonna use it on?
04:54:00 <klys_> yeah https://www.ebay.com/itm/305436666187
04:54:00 <bslsk05> www.ebay.com: VINTAGE DIGITAL DEC RX02-PA DISK SYSTEM FOR PDP-11 / PDP-8 Computer 1980s | eBay
04:54:00 <geist> i've thoughta bout trying to get the goteks working with mine
04:54:00 <klys_> this is for reverse engineering and 3d duplication
04:54:00 <geist> well done
04:54:00 <kazinsal> wonder if you can hook an RX02 up to a greaseweazle
04:55:00 <geist> iirc it's pretty standard, the pinout is different but otherwise it's pretty close to shugart standard or so
04:55:00 <klys_> yeah those are likely shugart
04:55:00 <geist> to the point that i have heard you can get goteks to work with older DEC stuff
04:55:00 <geist> they're not directly shugart, i bet they predate it, but i think it's fairly similar in the way the control flow work
04:57:00 <klys_> it might be worth noting I also have a pair of intellec 8" drives
04:57:00 <geist> that's a particular rabbit hole i haven't gone down yet
04:58:00 <geist> i've seen curiousmarc deal with HP 8" floppy drives too, i think the drives themselves were made by CDC or something
04:58:00 <klys_> hewlett packard or packard hell
04:59:00 <geist> the one true HP, back when their stuff was serious quality
05:01:00 <geist> wow, that's pretty cheap for an RX02 really
05:01:00 <geist> i guess the shipping might be a pain
05:02:00 <klys_> yeah was quite surprised
05:02:00 <geist> perhaps it's mostly because it was in a big case like that
05:02:00 <geist> so no one wants to deal with it
05:02:00 <geist> i dunno what that pin out is on the cable, but that's part of the fun
05:08:00 <klys_> seems I'd read some of this before too ttps://www.retrotechnology.com/herbs_stuff/drive.html
05:08:00 <bslsk05> www.retrotechnology.com: Floppy Drive Tech Info
05:09:00 <klys_> again the main idea is to duplicate the drive by making measurements
05:12:00 <klys_> naturally I can't seem to find my screwdrivers today
05:21:00 <klys_> there's some about the RX02: ttps://www.retrotechnology.com/herbs_stuff/m2fm.html#dec
05:21:00 <bslsk05> www.retrotechnology.com: M2FM or MMFM diskette format
05:29:00 <klys_> of couse MAME is still missing a rom from the decmate2 RX50 controller card rom (though I think the RX02 stuff should go there) except then I don't have that controller card, etc.
05:32:00 <klys_> https://github.com/mamedev/mame/blob/master/src/mame/dec/decmate2.cpp#L989
05:32:00 <bslsk05> github.com: mame/src/mame/dec/decmate2.cpp at master · mamedev/mame · GitHub
05:47:00 <klys_> https://www.tim-mann.org/trs80faq.html#[25] links to http://www.dbit.com/fdadap.html (explaining you should not plug this in to a pc) which might answer some of your question
05:47:00 <bslsk05> www.tim-mann.org: TRS-80 Model I/III/4: Frequently Asked Questions
05:47:00 <bslsk05> www.dbit.com: FDADAP floppy disk adapter
05:53:00 <klys_> and here's an adapter for sale, https://www.ebay.com/itm/276327034721
05:53:00 <bslsk05> www.ebay.com: TexElec's 34 to 50-Pin 8-Inch Floppy Adapter | eBay
05:59:00 <geist> ah yeah https://texelec.com/product/8-inch-floppy-adapter/
05:59:00 <bslsk05> texelec.com: 34 to 50-Pin 8-Inch Floppy Adapter - TexElec
05:59:00 <geist> thats the same guy that sells the commander x16
08:45:00 <mjg> 8"?
08:45:00 <mjg> i admit i never seen one in action :(
08:46:00 <mjg> i did grow up on 5.25" tho
09:11:00 <sskras> me too :)
09:18:00 <geist> Same, though I do have a lot of 5.25"s still
09:18:00 <geist> as much as I love retro computer tech though, floppy disks are just too finicky for me. I can't get over that every time you use one you're slowly destroying it
09:18:00 <geist> and old ones are just waiting to be destroyed next time you stick it in the drive
09:18:00 <GeDaMo> Isn't that true for flash too though?
09:19:00 <geist> yah but flash you can just get more :)
09:19:00 <geist> in 30 years or so yeah we'll be pretty upset that all the period SSDs are dead
09:20:00 <mjg> i don't know if people will give af about ssds
09:21:00 <geist> probably not, but when it all becomes retro even if it's not sexy now folks may be interested in it
09:21:00 <mjg> i mean in that very context
09:21:00 <geist> OTOH you can easily emulate floppies and hard drives now with flash, so in 30 years you'll just emulate the whole thing with whatever is the thing then
09:21:00 <mjg> there is always a limbo where nobody gives af
09:21:00 <geist> yeah
09:21:00 <geist> it's like a dead zone where 10 years on it's just old shit, and 20 years on it starts to get interesting
09:22:00 <geist> sort of an uncanny valley
09:23:00 <mjg> for me for example there 8-bit atari holds a sentimental value since it was my first computer
09:23:00 <geist> oh oh speaking of I just got one of the Atari 400 minis. it's really cute!
09:23:00 <mjg> apart from that maybe i could get some early 00s high end poc for some gaming, again sentimental
09:23:00 <mjg> but past that i have hard time imagining why i would give a fuck about a laptop i had later
09:24:00 <mjg> it's not a milestone "personally" or in the history of computing
09:24:00 <geist> yah that's why I had to get the precise 386 model I had as a kid when it showed up on ebay
09:24:00 <geist> it's not even a particularly good computer, but it was the exact model I had
09:24:00 <mjg> dude my first pc is still in the attic at my parents
09:25:00 <mjg> 486dx2 50mhz
09:25:00 <mjg> i got you beat on that aspect then :p
09:25:00 <geist> oh heck yeah keep that
09:25:00 <mjg> frankly though i could not be fucked to even check if it's still running
09:25:00 <geist> yeah I wished we haddnt junked the 386 years ago, but it was in the dark period when it was not particularly valuable
09:25:00 <geist> aww come on! that's nostalgia
09:25:00 <geist> next time you visit, get it out
09:26:00 <mjg> i have some nostalgia for *second* pc, a p75
09:26:00 <mjg> that 486 was total crap
09:26:00 <mjg> i only had 3 friends with pcs at the time and all of them had something faster
09:27:00 <mjg> including a different 486
09:27:00 <mjg> :x
09:27:00 <geist> yah I remember my 386sx 16 was a lot slower than my friends 386dx 20. of course it's obvious now why it was a lot faster but I had no idea what the diff was between sx and dx
09:27:00 <geist> I just wanted to dial into BBSes and get WAREZ and play games!
09:27:00 <mjg> and by faster i mean it was night and day
09:27:00 <mjg> fuckin' openbsd vs linux
09:28:00 <mjg> migrating to p75 was a massive jump
09:28:00 <mjg> i remember playing an mp3 took 100% of cpu time
09:28:00 <mjg> :d
09:29:00 <geist> heh yeah I remember playing a mp3 on a friends Packard bell 486. windows 3.11. I think it was the original fraunhaufer mp3 player
09:29:00 <geist> before winamp
09:29:00 <geist> and yeah pegged the cpu
09:29:00 <mjg> i think i "benchmarked" this later on some celeron 300, on that sucker it was about 25%?
09:29:00 <geist> that's about right. something like 50 MIPS to play an mp3
09:29:00 <mjg> this is where i learned about the power of code optimization tho
09:30:00 <geist> I remember looking at the ipod software about 2005 and at the time the ipad had a dual core arm7tdmi, and one of the cores was dedicated to decoding mp3/aac/etc
09:30:00 <mjg> i found a random-ass windows program to encode .wav into .mp3
09:30:00 <mjg> it would take forever
09:30:00 <mjg> then i found lame and fuck me, it was turbo fast in comparison
09:30:00 <geist> and they had it tuned so that it would change the frequency of the cpu to have just enough headroom to decode the file, base don the bitrate
09:30:00 <geist> and iirc it was something like 30-40mhz
09:32:00 <mjg> the one nice thing about that period is that any new cpu you got, everything got way faster
09:32:00 <geist> yah totally
09:32:00 <geist> late 90s early 2000s was the peak of it in my mind
09:32:00 <mjg> now webdev stay vigillant to make sure this does not happen
09:32:00 <geist> before SMP/SMT was starting to be the way
09:34:00 <geist> not that shit is not faster now, but up till about 2005 it was single core just doubling every 18 months and it was pretty exciting
09:35:00 <mjg> which years did you attend hs?
09:35:00 <mjg> did you have "computer science" or some such in there
09:35:00 <geist> 1991-1995, graduated in 95
09:35:00 <mjg> huh that's early enough to probably not?
09:36:00 <geist> yeah no, we had a computer class, but it was out of date even at the time. we had like first gen Compaq portable 286s (like 1985 tech)
09:36:00 <mjg> :P
09:36:00 <geist> it was laughable at the time. did some pascal programming
09:37:00 <geist> but I had ben programming some the 80s on my apple 2 and whatnot. BASIC and 6502 asm
09:37:00 <geist> honestly PCs were far less interesting to me at the time. I saw x86 as a piece of trash and no fun
09:37:00 <geist> I didn't get interested in os hackery until college, and a few years in at that
09:37:00 <geist> like '96 or so
09:37:00 <mjg> pc were always bullshit on that front
09:37:00 <geist> yah
09:39:00 <mjg> you know i got fucked over out of atari st as a kid? :(
09:39:00 <geist> oh dang
09:39:00 <mjg> ya
09:40:00 <geist> I went straight from apple 2 to PC
09:40:00 <geist> only later collected stuff like c64s and macs
09:40:00 <mjg> there was a dude who had 1040st with tons of hw for it
09:40:00 <mjg> and floppies
09:40:00 <mjg> and was looking to give it away
09:41:00 <mjg> i knew the guy in passing and he knew my grandma
09:41:00 <mjg> he wanted to give it to me
09:41:00 <mjg> but my grandma being anti-computer said it's bad
09:41:00 <mjg> so...
09:41:00 <geist> :(
09:41:00 <mjg> poland, 90s
09:42:00 <geist> I remember having the argument with my granddad about computers. he was a rancher in Texas and I remember going on about how computers ain't gonna raise cows, etc
09:42:00 <mjg> :d
09:42:00 <geist> which is actually silly, because really having a computer to track stuff would be useful
09:43:00 <mjg> my dad said computers are only good to do stuff in other fields, thus one needs a degree in, say, civil engineering
09:43:00 <mjg> and then one can make money using computers
09:44:00 <mjg> he also said there is no point going into computers anywya, the future is wellness industry
09:44:00 <mjg> i like to think about it when people claim parents always know best
09:44:00 <mjg> you are just too young to get it
09:44:00 <geist> heh yeah
09:46:00 <mjg> but if it so happened being in IT paid pooorly i would probably still be doing it anyway
09:46:00 <mjg> so there is that
09:47:00 <mjg> maybe overall code quality would be better as all the grifters would go elsewhere
10:09:00 <nikolapdp> heh
12:04:00 * Ermine gives gog a piece of cheese
12:05:00 * gog is fascinated
12:17:00 * leg7 notices there are worms growing inside the cheese
12:19:00 <Ermine> My cheese is worm-free
12:19:00 <leg7> https://en.wikipedia.org/wiki/Casu_martzu
12:21:00 <Ermine> I've heard about it, but I don't fascinate women with weird kinds of cheese
12:23:00 <zid> good news is that it's blackmarket only
12:23:00 <zid> eu regs don't allow its sale
15:53:00 <vin> How can I read the CR3 values? If I am to use rdmsr, where can I find the address?
15:53:00 <heat> mov %cr3, %rax
15:54:00 <heat> i dont understand the second question
15:54:00 <heat> note: mov %cr3, %rax does not work in userspace
15:56:00 <GeDaMo> https://sandpile.org/x86/msr.htm ?
15:56:00 <bslsk05> sandpile.org: sandpile.org -- x86 architecture -- model specific registers
15:57:00 <vin> heat: I would like to check the bit values of CR3[62] and CR3[61] from the userspace. This checks if LAM is enabled. If not enabled I want to update CR3[62] to 1 and keep CR3[61] to 0.
15:58:00 <heat> you can't do that
16:00:00 <heat> linux gives you some arch_prctl calls you can use
16:00:00 <vin> You mean I can't do that in the userpsace? Sec 6.1 of Intel Instruction reference says "LAM support by the processor is enumerated by the CPUID feature flag CPUID.(EAX=07H, ECX=01H):EAX.LAM[bitEnabling and configuration of LAM is controlled by the following new bits in control registers: CR3[62] (LAM_U48), CR3[61] (LAM_U57)"
16:00:00 <vin> I see
16:01:00 <heat> you can't touch cr3 from userspac
16:01:00 <heat> due to the obvious security vulnerabilities
16:02:00 <heat> like, i don't know why you're looking at the SDM, someone has carefully read all of that and carefully implemented it in the kernel
16:02:00 <heat> you just need to arch_prctl and ask for N tag bits, then use em
16:02:00 <heat> you have 0 control over any of this in userspace, except actually asking for tagged addresses
16:05:00 <vin> I see an example here https://github.com/llvm/llvm-project/blob/5c4b923c727767c9e6ad16324c559a1a92616171/compiler-rt/lib/hwasan/hwasan_linux.cpp#L139 but it isn't clear to me if this "enabled" LAM on the process and all addresses can have upper 15 bits available
16:05:00 <bslsk05> github.com: llvm-project/compiler-rt/lib/hwasan/hwasan_linux.cpp at 5c4b923c727767c9e6ad16324c559a1a92616171 · llvm/llvm-project · GitHub
16:07:00 <vin> I think this has a better example of enabling LAM https://github.com/ColinIanKing/linux-next/blob/f529a6d274b3b8c75899e949649d231298f30a32/tools/testing/selftests/x86/lam.c#L34
16:07:00 <bslsk05> github.com: linux-next/tools/testing/selftests/x86/lam.c at f529a6d274b3b8c75899e949649d231298f30a32 · ColinIanKing/linux-next · GitHub
16:07:00 <heat> yes it enabled LAM
16:08:00 <heat> that's a good example
16:08:00 <heat> IIRC fork'd processes also have the LAM enabled, but it'll clear off as soon as you execve
16:11:00 <vin> Makes sense, I noticed that LAM on intel is only available on Sierra Forest, Grand Ridge, Arrow Lake, Lunar Lake. I don't have access to any of them. Do people test LAM features using sde64 (emulator)?
16:14:00 <zid> They're tested by intel, then submitted to lkml
16:14:00 <zid> by intel
16:25:00 <vin> Yes testing is done by intel but if others want to icorporate the feature in their software, what do they do?
16:25:00 <zid> buy one
16:25:00 <zid> or get your OS to be something people use
16:26:00 <vin> zid: I don't think you can buy any server with the support yet
16:26:00 <zid> so then there's no problem
16:26:00 <zid> you have no customers to demand support from you
16:29:00 <heat> there's was a random qemu patch somewhere for LAM support
16:29:00 <heat> for tcg ofc
16:31:00 <heat> like, all these fancy features and whitepapers and all that jazz are published many years in advance of actual deployment
16:31:00 <heat> linux has supported PML5 since, what, 2016? it was first deployed around 2020
16:32:00 <zid> doesn't just apply to PCs
16:32:00 <vin> https://www.mail-archive.com/qemu-devel@nongnu.org/msg1015675.html ? I am not sure if this is just enablement (allowing qemu to use underlying LAM) or if it is emulation
16:32:00 <bslsk05> www.mail-archive.com: [PATCH v4 0/2] Add support for LAM in QEMU
16:32:00 <zid> think of like, usb specs
16:32:00 <zid> digital tv tuners, literally anything
16:32:00 <zid> most people only get to develop products *after* release
16:32:00 <zid> you have to be a big big player to get early access
16:32:00 <heat> vin, emulation
16:33:00 <heat> actually this isn't even emulation
16:33:00 <heat> just "recognize and passthrough the cpuid bit when functioning as a hypervisor"
16:34:00 <heat> zid, btw FRED is being merged this release i think
16:34:00 <heat> so FRED FRED FRED FRED in a couple of years i guess
16:34:00 <vin> yea heat. The only real emulation I saw this discussion https://community.intel.com/t5/Intel-ISA-Extensions/why-the-simulation-of-linear-address-masking-is-not-working/m-p/1578916 but haven't tested the older version yet.
16:34:00 <bslsk05> community.intel.com: why the simulation of linear address masking is not working? - Intel Community
16:34:00 <zid> oh shit
16:35:00 <zid> freeeed
16:35:00 <heat> there was an old patch emulating LAM on qemu
16:36:00 <heat> i did apply it once and i probably have the build somewhere in my drive
16:38:00 <vin> It will great if you can share the patch! heat
16:43:00 <heat> https://lore.kernel.org/linux-mm/20210205151631.43511-12-kirill.shutemov@linux.intel.com/
16:43:00 <bslsk05> lore.kernel.org: [QEMU] x86: Implement Linear Address Masking support - Kirill A. Shutemov
16:43:00 <heat> get a 2021 qemu version and try to apply it
16:46:00 <vin> Let me try that, thanks
16:49:00 <heat> zid, https://lwn.net/ml/linux-kernel/SA1PR11MB67347309287CC9207A6C1391A89A9@SA1PR11MB6734.namprd11.prod.outlook.com/
16:49:00 <bslsk05> lwn.net: RE: [PATCH v8 00/33] x86: enable FRED for x86-64 [LWN.net]
16:49:00 <heat> looks like they're ready to ship FRED hw
16:49:00 <heat> 1st half 2024 on their intel cloud whatever thing
16:58:00 <zid> fredfredfred
17:00:00 <heat> frederick
17:02:00 <sham1> freredick
17:30:00 <nikolapdp> heat was this the thing they want to use to skip the whole 16<->32<->64 transition thing during boot
17:31:00 <zid> that's x86s
17:32:00 <zid> FRED is fixing swapgs
17:32:00 <zid> They announced it around the same time I was implementing irqs so I was hype for it, and now it's actually coming, amaze
17:33:00 <zid> speaking of amaze, project hail mary is getting a movie adaptation
17:33:00 <zid> unsurprising after how well the martian did
17:33:00 <GeDaMo> The book's great
17:34:00 <zid> the book's amaze*
17:34:00 <GeDaMo> People keep saying that and I only vaguely remember it from the book :P
17:35:00 <zid> rocky never conjugated verbs properly (or rather, the synth)
17:35:00 <zid> so he called everything AMAZE
17:35:00 <zid> when's the next bobiverse book
17:35:00 <GeDaMo> I saw something about later this year
17:36:00 <zid> september 5th
17:36:00 <zid> kk
17:36:00 <zid> and has been optioned to universal apparently
17:56:00 <GeDaMo> A Rendezvous with Rama movie has been in development for more than 20 years :P
18:29:00 <kof673> > having a computer to track stuff would be useful farmers do in fact "use computers"
18:29:00 <kof673> you see similar things with stock traders with 6 monitors decrying the people who sit at a computer all day
18:29:00 <kof673> and drug dealers with their cell phones, those lazy computer people
18:30:00 <kof673> i would be the last person to say technology is magically "good" but much "opposition" is just an image thing, it is not seen as real work
18:30:00 <kof673> and if you think drug dealers and wall street has morals good luck to you lol
18:31:00 <kof673> besides, you need to learn cursive, because after elementary school that is what everything will be!
18:33:00 <kof673> in any case, this is arguably related to code quality: because some managers view people as interchangeable, throw away code, get to market, get the money, get in, get out
18:35:00 <kof673> it is a victim of its own success somewhat: computers are everywhere, roads are everywhere, why do we need construction workers? the problem is solved lol
18:37:00 <kof673> i'd say it is reverse bikeshed phenomenon in a way: things appear to be "solved" and looks to be simple/easy, anyone can do it, etc.
18:39:00 <kof673> if one only interacts with a final end result, then all the work that went into that is not visible, and it is just people dicking around clicking things all day, gamers lol
18:40:00 <kof673> why do we need more people who just play candy crush all day etc. lol
18:42:00 <zid> hey that's looking like good news though
18:42:00 <zid> villenveuve is reportedly actually writing a script for it
18:42:00 <zid> after 20 years of morgan freeman going "yea I wanna make it, just need a script"
18:45:00 <kof673> lol sorry for length i just mean people will literally badmouth "computer people" (health, banking, other) as they literally are using a computer depending on it for their livelihood lol
18:48:00 <geist> huh, villeneuve is writing a script for rama?
18:49:00 <geist> i thought he was working on the next dune movie
18:49:00 <zid> He's directing that right?
18:49:00 <zid> I assume he gets a lot of work hours that are empty on top of directing a movie for the couple of weeks the shoot actually happens
18:50:00 <zid> (plus a couple of weeks of meetings, and some editing and press and stuff)
18:52:00 <geist> i seriously doubt that. he sounds like a serious perfectionist
18:52:00 <zid> well a shoot only lasts so long
18:52:00 <zid> it's expensive and the producers organize it to last the least amount of time possible
18:52:00 <zid> equipment rental etc over the least possible time
18:52:00 <GeDaMo> Possibly he spends a lot of time travelling
18:53:00 <zid> Yea I imagine most of his job is actually meetings with studio execs :P
18:58:00 <kof673> mjg: i am not disagreeing with "webdev" but in a post-agile world j/k where supposedly the "stakeholders" all supposedly meet and design products together...why do "bad" projects get funded? why do managers choose "bad" projects? why do "consumers" and "clients" ? the dev is only one tiny little part of the picture
18:59:00 <kof673> you need never hear other lines of work say this "noone has to buy, noone has to sell"
18:59:00 <kof673> *you would
19:01:00 <kof673> why do school churn out "web devs"? etc.
19:01:00 <kof673> *schools it seems a way for all the other players to avoid looking in the mirror
19:09:00 <kof673> socialize the profits, outsource the responsibility lol
19:21:00 <kof673> *socialize/capitalize in any case, all the other "stakeholders" have vanished when accountability is summoned lol
19:29:00 <heat> geist, IMPLEMENT FRED!!!!
19:30:00 <geist> oh yeah?
19:30:00 <geist> did fred get rolled out?
19:30:00 <heat> it's getting rolled out
19:30:00 <heat> they're merging the FRED code for this linux release
19:31:00 <heat> https://lwn.net/ml/linux-kernel/SA1PR11MB67347309287CC9207A6C1391A89A9@SA1PR11MB6734.namprd11.prod.outlook.com/
19:31:00 <bslsk05> lwn.net: RE: [PATCH v8 00/33] x86: enable FRED for x86-64 [LWN.net]
19:31:00 <heat> it's Coming Soon(tm)
19:32:00 <zid> fredfredfredfred
19:35:00 <puck> goes well with the asrock rack PAUL and TOMMY and
19:35:00 <puck> s/and $//
19:36:00 <zid> I prefer huawei extended attribute tables
19:36:00 <zid> and z80 index descriptors
19:37:00 <heat> puck, hey you should enable the bslsk05 sed thingy
19:37:00 <puck> maybe yeah
19:40:00 <geist> neat. i wonder what intel cores are getting fred first
19:41:00 <heat> xeon probably, then a few generations later it comes to desktop and laptop
19:41:00 <heat> it's how they did PML5
19:43:00 <zid> no no, it will be in *half* the platinum xeons
19:43:00 <zid> but all of the bronze ones
19:43:00 <zid> then in half the gold, but only 1 platinum, and none of the silver or bronze
19:43:00 <zid> then next gen it will be in the 18800kf
19:43:00 <zid> but nothing else
19:43:00 <zid> oh wait no this is avx isn't it
19:48:00 <mjg> where does amd stand in this
19:52:00 <childlikempress> they'll support it across the lineup but half the functionality will be microcoded and super slow
19:53:00 <zid> ooh even better
19:53:00 <zid> I hope we get FREDv2 etc as cpuflags soon
19:54:00 <zid> because of a bunch of exploits, and whether it's fast enough to use etc
19:54:00 <zid> we're already up to avx10.1
19:54:00 <zid> and avx10 isn't even out
19:55:00 <geist> iirc AMD had some sort of competing idea out, but it was far less ambitious than FRED
19:55:00 <geist> iirc FRED was nice, but had some fatal flaw for fuchsia that i dunno how to deal with
19:55:00 <heat> oh yeah?
19:55:00 <geist> i think iirc it removes the io bitmap
19:55:00 <heat> oh
19:55:00 <heat> yeah probably removes the tss?
19:55:00 <geist> since there's no such TSS equivalent
19:55:00 <heat> right
19:56:00 <geist> i dont think we use it much, but it's possible for a user driver to need it if it needs to use an io range
19:56:00 <geist> which is fairly far and few between, but legacy stuff like ps2 keyboards and whatnot...
19:57:00 <geist> i wonder when qemu will get fred. i'll probably fiddle with it there first
19:58:00 <geist> havne't really internalized how easy it is to support fred and not fred in the same build. or is it different enough that it's cleaner/easier to make a new sub-arch of x86
19:58:00 <mjg> would you even care in fuchsia to support both?
19:58:00 <heat> geist, could you force all outb/w/l (and in) to a syscall?
19:59:00 <geist> probalby would have to do that, which is clearly lame.
19:59:00 <geist> but i'm sure that's the worst case answer
19:59:00 <heat> or do some nasty x86 fault emulation by looking at the instruction
19:59:00 <heat> that would be EPIC
19:59:00 <geist> yah sadly trapping from ring3 you dont really get all the assist you get when trapping via a vmexit on intel and amd
19:59:00 <geist> unless.... fred does that?
20:00:00 <geist> if it decoded it for you and put it in the fred exception then that might not be so bad
20:00:00 <heat> who knows
20:00:00 <heat> that'd be cool
20:00:00 <heat> but i bet the port io insns don't have that many forms
20:00:00 <heat> should be okay to interpret manually
20:00:00 <geist> right. iirc intel and amd hypervisor traps have basically full decode assists for the io isntructions
20:00:00 <geist> you really only need to know which ones they are, and if it's an io with an #imm8 what the immediate is
20:01:00 <geist> otherwise all the info is in eax and ecx (or is it edx?
20:01:00 <geist> )
20:02:00 <heat> i'd love if FRED was a huge exception rework but, from the little of the patches i read, it seems like it keeps the old 32 exception vectors and everything
20:02:00 <heat> i'd love if GPF could get broken down into sensible-er exceptions
20:04:00 <zid> fred doesn't go too far enough yea
21:07:00 <zid> honzuki is laaaate, save me