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=7&d=4
00:00:00 <Matt|home> o\
00:08:00 <mjg> gog: oy
00:08:00 <mjg> how old of a laptop
00:40:00 <gog> 2 years
00:40:00 <nikolapdp> which laptop is it anyway
00:41:00 <mjg> whose
00:41:00 <gog> https://pcsupport.lenovo.com/is/en/products/laptops-and-netbooks/gaming-series/ideapad-gaming-3-15arh05/82ey/82ey000vmx/pf32yqwm?linkTrack=Homepage%3ABody_Search%20Products&searchType=6&keyWordSearch=PF32YQWM
00:41:00 <bslsk05> pcsupport.lenovo.com: laptops and netbooks :: gaming series :: ideapad gaming 3 15arh05 :: 82ey :: 82ey000vmx :: pf32yqwm Lenovo PC Support - Lenovo Support IS
00:43:00 <nikolapdp> kind of similar to mine, except mine doesn't have a dedicated gpu
00:44:00 <gog> the keyboard has this fucky weirdness about it which it hasn't been exhibiting for the last few days
00:44:00 <gog> idk what it's about
00:44:00 <gog> but the s key was hit or miss for awhile, then it was sending multiples, now it's fine
00:45:00 <nikolapdp> odd
00:45:00 <gog> yes
00:45:00 <nikolapdp> and that's what happened when it was new
00:46:00 <nikolapdp> ?
00:55:00 <zid> My keyboard likes to randomly insert a different character sometimes
00:55:00 <nikolapdp> are you sure it's not your parkinsons'
00:55:00 <zid> 1/10000 chance that you get a 9 instead of a u or something
01:09:00 <heat> mjg, those readlinks, are they disambiguating -l switches from obj files?
01:09:00 <heat> and if so, why readlink and not access or stat
01:29:00 <adder> how am I supposed to stay safe when my security guard keeps falling asleep https://arcticadder.net/acab1312/photo.jpeg
01:32:00 <nikolapdp> i don't think your link works
01:34:00 <nikolapdp> kek
01:34:00 <Mutabah> `photo.jpeg: ISO Media, HEIF Image HEVC Main or Main Still Picture Profile`
01:35:00 <nikolapdp> well firefox didn't like it
01:35:00 <Mutabah> Yep, but nomacs liked it - d'aww
01:36:00 <adder> this one should work https://arcticadder.net/acab1312/photo_2024-07-04%2003.35.59.jpeg
01:36:00 <adder> fakin epl
01:37:00 <nikolapdp> epl?
01:37:00 <adder> apple
01:37:00 <nikolapdp> of course
05:47:00 <Mondenkind> wait a minute are you allowed to do non-lock-free c11 atomics in signal handlers
05:48:00 <Mondenkind> i never even thought about this
06:57:00 <mjg> Mondenkind: ?
06:57:00 <mjg> how tf would that work
06:57:00 <Mondenkind> i mean i can imagine bad ways of making it work
19:54:00 * vdamewood sniffs gog
20:45:00 <nikolar> Are there systems with cache line sizes >16 in common use
20:45:00 <Mondenkind> >16 what
20:45:00 <nikolar> Erm bytes, my bad
20:46:00 <Mondenkind> yes
20:46:00 <nikolar> Do you know any
20:46:00 <Mondenkind> i'm typing on one right now
20:46:00 <Mondenkind> and i expect you are too
20:46:00 <Mondenkind> usually like 64b
20:47:00 <nikolar> Are there any with like 256 or 512
20:47:00 <Mondenkind> i don't think so
20:48:00 <nikolar> I've read somewhere that there are but it sounds like a bit much
20:48:00 <zid> I imagine that'd work great on supercomputery vliwy stuff
20:49:00 <zid> like, you wouldn't care if it was 512B if you're exclusively running avx-512 streamy stuff
20:50:00 <nikolar> Guess so
20:50:00 <nikolar> But I doubt that server processors have that much
20:50:00 <nikolar> Like it's the same microarch
21:10:00 <geist> nikolar: pretty much all are like 32 or 64
21:10:00 <geist> i've seen 128 byte cache lines (thunder x1 cpus) but i think the standard is 64 for the most part
21:10:00 <geist> i've also seen on i thinnk some of the x86 cpus having an L2 or L3 cache hae double the cache line size as the L1
21:11:00 <geist> which makes sense, caches farther out could easily deal with that
21:11:00 <Mondenkind> i've not seen that, more usually 'adjacent line prefetch'
21:11:00 <geist> yeah i forget which one i saw it on
21:11:00 <Mondenkind> so if you fetch a 64b cache line its buddy comes along for the ride into l2/l3 (and not into l1) but then it can get kicked out separately later
21:11:00 <Mondenkind> i think apple cpus do that though with the two line sizes
21:12:00 <geist> possibly yeah
21:12:00 <geist> also there's a trend of these distant whole-memory caches
21:12:00 <Mondenkind> huh?
21:12:00 <geist> different names for it, but kinda like a cache at the DRAM controller, so that it's totally invisible to everything
21:13:00 <geist> i dunno what the correct compsci name is, but i've seen it on a lot of the newer high end ARM socs, including the apple M1+
21:13:00 <geist> it's kinda like a L4, but basically completely invisible to anything, including DMA, so you can logically think of it as a cache on the DRAM controller
21:14:00 <Mondenkind> hrm. makes sense i guess?
21:14:00 <geist> SLC (system level cache) i've seen thrown around
21:14:00 <mjg> is there a context to that 16 byte question tho
21:14:00 <geist> yah especially for SOCs where the gpu and cpu and all these devices are sharing the same memory
21:14:00 <mjg> pretty weird as is
21:14:00 <Mondenkind> i wonder if it's for power more than speed
21:15:00 <geist> very possibly
21:15:00 <Mondenkind> no that doesn't make sense
21:15:00 <geist> apple cpus do it, i've seen it on samsung and qualcomm and google's smartphone cpu
21:15:00 <geist> seems to just be the defacto solution for high end smartphone/tablet class ARM socs now
21:16:00 <geist> anyway in that case the cache line could also be whatever size it wants to be for implementation purposes
21:16:00 <geist> since for the most part you'd probably only need to flush or invalidate it around power events
21:19:00 <Ermine> tlp somewhy switched me from intel_pstate to intel_cpufreq
21:21:00 <Ermine> ... bacause i chose passive opmode
21:27:00 <nikolar> mjg basically what's the required aligned to avoid false sharing
21:31:00 <mjg> depends, by default you roll with 64 within structs
21:32:00 <mjg> and 128 for globals
21:34:00 <nikolar> That's it yeah
21:36:00 <Ermine> PML5 PML5 PML5 https://www.phoronix.com/news/Linux-Unconditional-5-Level-x86
21:36:00 <bslsk05> www.phoronix.com: Linux Looking To Make 5-Level Paging Support Unconditional For x86_64 Kernel Builds - Phoronix
21:37:00 <zid> I appreciate that approach in general, but it seems a little.. overkill?
21:37:00 <zid> it's not like removing the option removes any code
21:37:00 <zid> just stops pml4 only builds existing
21:38:00 <zid> you still need the pml4 only codepath, you still need the pml5 codepath, you still need a bunch of runtime checks for which to use
21:38:00 <nikolar> Does that mean that all kernels will work with both?
21:38:00 <nikolar> Yeah
21:39:00 <geist> for what it's worth most modern arches at least either have it hard coded or a way to query it
21:40:00 <Ermine> I guess less #ifdef CONFIG_ clutter
21:41:00 <mjg> i'm guessing that's mostly to reduce maintenance burden
21:41:00 <mjg> not for any gainz for anyone
21:41:00 <geist> yeah it's not that they're forcing 5 level, just make it so that it can choose to use it by default
21:41:00 <geist> i'm guessing there's probably a command line switch to force it back to 4 even if the hardware supports 5
21:41:00 <geist> in case there's some slight performance delta, etc
21:42:00 <geist> stupid intel for not just making it per aspace, like ARM
21:42:00 <geist> even FRED doesn't fix that
21:42:00 <mjg> there very well may be dragons in the silicon
21:43:00 <geist> oh almost certainly. the downsides of having so much legacy, there are probably whole blocks of verilog they dont want to change, etc
21:43:00 <geist> so there would be a very strong reason to not futz with it
21:43:00 <geist> practically speaking
21:43:00 <mjg> ye
21:43:00 <mjg> realistically the 5th level iso for such a miniscule amount of people
21:44:00 <mjg> it does not make a difference
21:44:00 <geist> would be interesting to see what the performance delta may be for an additional PTE level. depends on TLB miss hit rate and how well the PT walker cache works
21:44:00 <geist> which is mostly invisible on both intel and amd, though clearly present
21:44:00 <geist> also obviously depends on the clustering of the test cases, etc
21:44:00 <geist> would have to be a Real World test
21:45:00 <mjg> fuck that, just declare there is no difference
21:45:00 <mjg> it's i/o bound to the network
21:45:00 <geist> i assume you jest, considering how much your'e about benchmarks
21:46:00 <mjg> dude my rants about geezers which take shit out of their asses take second place after rants about solaris
21:46:00 <mjg> which btw is also turbo geezered
21:46:00 <geist> okay,so it was sarcasm
21:46:00 <mjg> i'm also more about profiling than benchmarks,
21:46:00 <geist> can't tell with you
21:46:00 <mjg> :sadface:
21:47:00 <geist> tis why i try to make it obvious or put a sarcasm tag or whatnot. lots of it gets lost in IRC and translation
21:47:00 <geist> i guess discord you'd put a gif afterwards that uses up half the page with some cute thing
21:47:00 <zid> my kingdom hearts crashed :(
21:47:00 <geist> eep
21:47:00 <mjg> :shieeet:
21:48:00 <geist> i reinstalled my main home server with proxmox yesterday and re-set up a large VM that acts like the old box's native OS
21:48:00 <geist> working pretty well, i gotta admit it's a pretty slick system for free
21:49:00 <Ermine> I stumble upon proxmox when googling for lxc stuff pretty often
22:08:00 <geistpdp> nice, still connected
22:08:00 <geist> yeah it has support for lxc
22:53:00 <geistpdp> time to shut the pdp down, let it rest a bit
22:54:00 <geistpdp> /me waves
22:54:00 <geistpdp> (interesting, it doesn't do /me)
22:55:00 <nikolapdp> bye
22:55:00 <geist> but shows you how stable it is on real hardware
22:55:00 <geist> ran for a few days with no trouble
22:56:00 <geist> it's loud though, sometimes it get son your nerves, wanted to shut it down to get my hearing back
23:01:00 <nikolapdp> nice to see it working
23:01:00 <nikolapdp> did you sort out the line wrap issue
23:04:00 <mjg> e10k in the basement when
23:12:00 <geist> surprisngly it doesn't burn that much power, only about 80W
23:12:00 <geist> but i bet a solid 20 of it goes into the old style AC fans
23:13:00 <nikolapdp> heh neat
23:13:00 <nikolapdp> i really want to get one, at least one of the small ones but they are basically unobtainable over here
23:13:00 <nortti> geist: how reasonable would it be to mod it to use 12V fans?
23:21:00 <geist> oh it'd probably be fine, but actually it is shoving a lot of air through it, so i think you'd need some fairly beefy fans to get the same effect
23:21:00 <geist> part of the noise is just the amount of air it's moving, it's a rack mounted thing ultimately, though it's shoving the air through sideways FWIW
23:22:00 <geist> i took it apart at some point and re greezed their bearings but i didn't take any pics
23:22:00 <geist> the power supply is a monster and the fans are part of that assembly
23:22:00 <zid> geist your accent is showing
23:22:00 <geist> oh yeah?
23:23:00 <zid> regreased :p
23:23:00 <nikolar> lel
23:23:00 <geist> ah. nah i just didn't bother trying to spell it right
23:23:00 <zid> right but you only spell it with a z if you have that accent that says it with a z, surely
23:23:00 <geist> but yeah. you're right. honestly 'grease' as a word looks a bit off for me
23:23:00 <geist> oh i definitely do
23:23:00 <zid> so your accent was infact, showing
23:24:00 <geist> that's probably why it looks off to me, since i say 'gree-z'
23:24:00 <zid> I have the opposite with 'houses'
23:24:00 <geist> yah i'd say 'how-zez'
23:24:00 <zid> I say houzes and greased, I've seen americans who say houses and greazezd
23:24:00 <zid> I should find one and duel him to the death
23:25:00 <geist> i think that's a bit different from the local seattle proper accent which would be more of a proper houses saying
23:25:00 <geist> i bet the s -> z thing is the overall summary and some southern thing
23:26:00 <geist> just took the spark plug out of my vespa trying to find a reacplement online
23:26:00 <geist> geez there's like 15 different variants
23:26:00 <geist> no standard it seems for spark plugs
23:26:00 <zid> Just look up what the oem gave for the specific model of vespa you have?
23:27:00 <zid> as long as it has the right thread pattern and spark gap etc it'll be fine though
23:27:00 <geist> yah i have a champion RG6YC in it, which i can always order
23:27:00 <geist> though it's basically the cheapest thing on the market
23:27:00 <geist> looks like a NGK CR8EIB is a solid replacement and higher quality
23:28:00 <zid> what have you been doing to it such that you've been eating spark plugs though
23:28:00 <geist> honestly nothing, but it's got like 6k miles and 15 years on it, so i think it's time
23:28:00 <geist> i dot think it's fouled, but it's not running well lately, been missing somewhat and whatnot
23:28:00 <geist> so i pulled it out. it doesn't seem to have a lot of carbon buildup, but figured what the hey
23:29:00 <zid> you can just scrape the fouling off
23:29:00 <zid> as long as the electrode isn't super worn down
23:29:00 <zid> is it a 2 stroke?
23:29:00 <zid> I imagine those foul a lot quicker
23:30:00 <geist> https://usercontent.irccloud-cdn.com/file/nvZMNTfF/sparkplug.pmg
23:30:00 <zid> oh that's fiiine
23:30:00 <geist> frankly it's a bitch to get it out of the scooter
23:30:00 <geist> you think?
23:30:00 <zid> but if it's out now, might as well change it?
23:31:00 <zid> stops you having to do it again
23:31:00 <geist> basically
23:31:00 <geist> that was my idea, if i can get a new one in the next day or two may as well just replace it
23:31:00 <zid> It's the same with engine gaskets and stuff
23:32:00 <zid> they're cheap and consumable, so if you ever have the engine apart
23:32:00 <zid> you might as well replace everything
23:32:00 <geist> the plug is a PITA to get to too, you need small hands and lots of blind reaching into the bike cavity to get it
23:32:00 <zid> did you not have a good socket for it?
23:32:00 <geist> thankfulyl they included a tool with the bike, no way to get a socket anywhere near it
23:33:00 <zid> https://m.media-amazon.com/images/I/61FriAcHSqS.jpg
23:33:00 <geist> no way a socket works. but they included a little socket like thing with some holes that you can stick a screw driver in to turn it a quarter
23:34:00 <geist> https://youtu.be/3vbeJ4bgvPA?t=101
23:34:00 <bslsk05> 'Vespa LX Spark Plug Replacement Tutorial | Mitch's Scooter Stuff' by Mitch's Scooter Stuff (00:04:29)
23:34:00 <zid> yea you'd need a very specific shape
23:34:00 <zid> but a socket tool *would* work
23:34:00 <geist> a very shallow socket might do it
23:34:00 <zid> one that could go through 90
23:34:00 <geist> right
23:35:00 <geist> but whatever. thinking about replacing it anyway
23:35:00 <zid> yea do it
23:35:00 <geist> it's taken to stalling out when accellerating, but only when it's warm as far as i can tell
23:35:00 <zid> they're only a few dollars I assume
23:35:00 <geist> yah about $10
23:35:00 <zid> check fuel filter
23:35:00 <geist> yep. also something that is a terrible pita
23:36:00 <zid> I believe the technical term for that is
23:36:00 <zid> "has an engine"
23:36:00 <geist> well, the main problem is it's so jamed up underneath it. it's really designed to work on with a proper table lift
23:38:00 <zid> If you don't have to disassemble the entire thing to do it, it's considered 'very accessable'
23:38:00 <zid> https://fxtwitter.com/jackmeridan/status/1807943650204381451?s=46&t=PsZzK1JHVJ4Cadv3I7rIpA
23:38:00 <bslsk05> twitter: <jackmeridan> The most painful thing you will see on here today. https://video.twimg.com/ext_tw_video/1807943590020370434/pu/vid/avc1/576x1024/H-clvaXohfunVvy9.mp4?tag=12
23:39:00 <geist> i think he could have summed it up with a picture of a mini in the shop
23:41:00 <zid> I assume this is a joke video though and they needed that engine apart anyway
23:41:00 <zid> otherwise he'd have just fished it out with a magnet
23:46:00 <geist> ordered a nicer, higher end iridium plug that folks seem to like
23:46:00 <zid> if you wanna test that plug btw you can just attach it to the HT lead without screwing it in
23:46:00 <zid> and see if it sparks when you turn the engine over
23:47:00 <geist> yah though i dont think i have enough arms to do that in this case
23:47:00 <zid> have you considered buying a few more? might be worth it in the long run
23:50:00 <geist> reminds me of project binky
23:50:00 <zid> Good news btw, today is the last day of some americans having all of their fingers and arms
23:50:00 <zid> you can probably get some extras cheaply
23:50:00 <geist> https://www.youtube.com/watch?v=7hCPODjJO7s&list=PLGSOZAHg1yQHU1tc_3Y5MTQg1qjtxA_nq&index=1
23:50:00 <bslsk05> www.youtube.com <no title>
23:51:00 <zid> Nobody mentioned brackets though
23:51:00 <zid> or CAD