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=22&m=6&d=11

Saturday, 11 June 2022

00:47:00 <heat> macbook users
00:47:00 <heat> how tf do the thermals work
00:47:00 <heat> my macbook pro (intel, 2019?) is doing 100C *easy*
00:48:00 <citrons> I'm not a macbook user, but the answer is that they don't. many macbook models have problems that often arise from them basically melting themselves
00:49:00 <zid> They're designed to turbo, saturate the heatsinks, throttle
00:49:00 <zid> then wait for the next time you want to turbo
00:49:00 <mrvn> you don't use a mac book the mac book uses you
00:49:00 <moon-child> macbook pro (m1 pro, 2021): ice cold
00:49:00 <moon-child> :P
00:49:00 <zid> You can get a couple seconds of full turbo if you want!
00:49:00 <heat> like the fan doesn't even kick in at around 70-80C
00:50:00 <moon-child> I think fan noise is contrary to '''the apple aesthetic''', so they tune it to be conservative. Annoying
00:50:00 <mrvn> I want transistors that work at higher temps so 100°C at the heat sink is normal and keeps the tea hot.
00:50:00 <zid> Wanna see some amazing code
00:51:00 <moon-child> sure
00:51:00 <zid> https://github.com/BLNJ/JADE/blob/master/JADE.Core/CentralProcessingUnit/InstructionManager.cs
00:51:00 <bslsk05> ​github.com: JADE/InstructionManager.cs at master · BLNJ/JADE · GitHub
00:51:00 <mrvn> heat: if they have good heat transmission then 80°C is not an issue
00:51:00 <zid> They've started to write a gameboy emulator, but it's in enterprise java style, completely unironically
00:51:00 <heat> late stage OO
00:51:00 <zid> they got angry at me when I joked about it
00:51:00 <moon-child> wow
00:51:00 <zid> They've reimplemented arrays about 6 times in here via dispatch method stuff
00:51:00 <heat> mrvn, i'm getting die temps
00:51:00 <heat> supposedly
00:52:00 <heat> but like
00:52:00 <heat> how can you ever have this on your lap?
00:52:00 <zid> Turns out if your laptops are 3mm thick
00:52:00 <heat> you'd be sterile before the compiltions ends
00:52:00 <zid> you can't have a 100W cpu and have it not throttle after a nanosecond
00:52:00 <heat> compilation*
00:52:00 <zid> nobody does compilation on a macbook, they do TRANSPILING TO JAVASCRIPT
00:53:00 <zid> but in a docker
00:53:00 <zid> on a machine they're a thinclient for
00:53:00 <heat> i'm doing that
00:53:00 <heat> how about that
00:53:00 <moon-child> I compiled c on a macbook recently
00:53:00 <moon-child> with make
00:53:00 <zid> did you live?
00:53:00 <heat> it's on my desk and the radiation hasn't fukushima'd me yet
00:53:00 <moon-child> barely
00:54:00 <zid> fukushima was a spent fuel pool with a leak, meh
00:58:00 <heat> moon-child, i'm compiling c++
00:58:00 <heat> it's literally 2 steps away from blowing up :P
00:58:00 <moon-child> that seems to be the stable state for most c++ code
00:58:00 <zid> Compiling C++ is genuinely a threat to human health
00:58:00 <zid> unlike fukushima
00:59:00 <mrvn> I should compile some c++, it's cold.
01:03:00 <heat> i genuinely don't get how they think its ok to operate at a temp where I could *boil fucking water on the die*
01:03:00 <heat> ?????????
01:04:00 <heat> do people spend 2200 euro on a laptop just to use it for google docs
01:04:00 <heat> if i touch the shell for too long I'll get 3rd degree burns
01:05:00 <zid> yes.
01:05:00 <mrvn> 100° on the die is nothing. the shell temp is an issue though
01:05:00 <mrvn> but laptops that brn your knees is nothing new.
01:05:00 <zid> there are people who literally do their jobs via.. telephones
01:05:00 <zid> they don't even use a computer AT ALL :(
01:05:00 <heat> how is 100C on the die nothing?
01:06:00 <mrvn> because it's still way below failure temp
01:08:00 <citrons> I think my computer will shut itself off at temperatures higher than 100C
01:08:00 <heat> mrvn, the tjunction for the cpu is supposedly 100C
01:09:00 <mrvn> citrons: it depends where you measure
01:10:00 <mrvn> https://en.wikipedia.org/wiki/Wide-bandgap_semiconductor
01:10:00 <bslsk05> ​en.wikipedia.org: Wide-bandgap semiconductor - Wikipedia
01:10:00 <mrvn> "allowing them to operate at much higher temperatures on the order of 300 °C."
01:10:00 <mrvn> Now make me a CPU of those so I can turbo the shit out of it
01:11:00 <citrons> at some point, it makes more sense to rate it in BTUs because it's more effective as a space heater
01:12:00 <mrvn> 80W is 80W no matter how you measure it
01:13:00 <mrvn> 300°C semiconductors would be nice to stack more ram on a CPU core for a SOC.
01:14:00 <clever> where would you draw the line between soc and som?
01:14:00 <mrvn> I wonder why noone has designed dies with holes through them that you can fill with copper or pipe heat fluid through.
01:14:00 <mrvn> clever: I don't care. I just want more ram on my RPi.
01:15:00 <clever> for the vc4 lineup, the hw design has a max of 1gig of ram, and it must present itself over a single 32bit ddr2 bus
01:15:00 <clever> and nobody actually makes 1gig dram dies
01:16:00 <clever> the 1gig model of the pi, is infact using a pair of 512mb dies, each on half of the 32bit bus
01:16:00 <clever> so every 32bit word in ram, is effectively striped over 2 ram dies!
01:17:00 <clever> for the bcm2711, the hw can address up to 16gig of ram, but the same problem exists, nobody makes a single-die 16gig
01:17:00 <clever> and 8gig is the best you can get currently
01:17:00 <clever> to go beyond that, you either need an soc that with more dram busses, like x86
01:17:00 <clever> or you need fatter dram chips
01:18:00 <mrvn> clever: The problem is you can't stack dies without end because the heat has to travel through all of them to reach the heat sink.
01:18:00 <clever> yeah
01:19:00 <mrvn> But if the die would run fine at 300°C you could stack that much more dies.
01:19:00 <_Heat> if I see you referencing reserved identifiers i'll be reporting it to the C standards committee
01:20:00 <clever> isnt the apple M1 not stacking, but just placing all of the dies on a much larger substrate?
01:20:00 <mrvn> given the size of the thing that wouldn't surprise me
01:21:00 <clever> the rpi zero-2 is a special case
01:21:00 <clever> the design goal, was to make a single-sided pcb
01:22:00 <clever> when using the bcm2835 and package-on-package, that was easy
01:22:00 <mrvn> I wonder how much better you could do multiplication, division, sqrt, trig functions if you build your gates truely 3D with say 256 layers.
01:23:00 <clever> another factor at play, is that the bcm2835 was designed for both wire-bonding and package on package
01:23:00 <_Heat> https://www.youtube.com/watch?v=5TFDG-y-EHs
01:23:00 <bslsk05> ​'NaN Gates and Flip FLOPS' by suckerpinch (00:19:07)
01:23:00 <clever> so the locations of the dram pins on the raw die, was designed to match up to the dram pins on a raw ram die
01:23:00 <_Heat> this is peak hardware eng
01:23:00 <clever> but the bcm2837 wasnt designed with that in mind
01:24:00 <clever> so the pi-zero-2 is pulling off some crazy hacks to even do it
01:24:00 <clever> basically, the data bus for the ram, is totally mis-wired
01:24:00 <clever> bit0 of the ram, is wired to bit5 of the controller, for example
01:24:00 <clever> but like the enigma machine, all reads go thru that rats-nest in the reverse direction of writes, and it unscrambles
01:26:00 <clever> mrvn: but, they didnt/couldnt jam 2 ram and 1 soc into the same package
01:27:00 <_Heat> all pi's are enigma machines in the sense that it's an enigma to find out why things were designed in the way they were
01:31:00 <mrvn> heat: using bool = double?
01:33:00 <heat> no
01:33:00 <heat> long double at least
01:33:00 <heat> you need the precision
01:37:00 <mrvn> "on a balesian 2 dollar node -- for scale." Sure, we all know how big those are.
01:38:00 <mrvn> note even
05:07:00 <geist> hah cute: i downloaded from ESP32s site the dev software for their esp32-c3
05:07:00 <geist> and it's a 800MB .zip file
05:07:00 <geist> inside the zip file is the git repository that you can check out
05:07:00 <geist> including the .git pack files and everything
05:08:00 <geist> in fact like 750MB of the 800MB is the compressed .git/ dir
05:13:00 <kazinsal> whoops!
05:55:00 <zid> fun, M1's pointer tagging got broken
05:56:00 <zid> speculatively fetching from tagged pointers to avoid actually having to deref them and crashing
05:58:00 <zid> eventually you get a kernel TLB hit and decode the correct kernel tag
06:11:00 <mrvn> Youz just try tags till you get one that's faster?
06:23:00 <geist> hmm, what precisely do you mean by tagged pointers?
06:25:00 <mrvn> one of the extensions where you use the upper bits of addresses to verify the pointer.
06:26:00 <kazinsal> having flashbacks to the idea of "32-bit clean/dirty" in system 7
06:26:00 <zid> PAC
06:27:00 <geist> ah i was just surprised that M1 implemented that
06:27:00 <zid> It seems like a neat idea overall
06:27:00 <zid> but we live in a post spectre world so "oops" in terms of implementation
06:28:00 <geist> we definitely can't have nice things
06:28:00 <geist> boooo
06:29:00 <kazinsal> it's fascinating that apple would implement something like that
06:29:00 <zid> Just stop allowing high precision timing in userspace, easy
06:29:00 <zid> (that's how javascript fixed it heh)
06:29:00 <kazinsal> considering having previously been bit by "use the currently unused bits in pointers for storing handle data"
06:31:00 <geist> yah agreed
06:32:00 <mrvn> here is a nice mitigation: run opcodes in constant time and use cached results to allow the core to cool down.
07:04:00 <mrvn> Anyone know of a multi precision bignum library that uses SIMD and multithreading?
07:06:00 <moon-child> why do you want that?
07:07:00 <mrvn> bored
07:08:00 <moon-child> lots of dependencies. Parallelisation is hard
07:08:00 <moon-child> I think libbf uses simd for its fft multiply
07:08:00 <moon-child> gmp obviously does all the things
07:08:00 <moon-child> addition is _technically_ logn span in parallel, but that requires a tonne of redundant work so it doesn't really make sense to do
07:10:00 <mrvn> The gmp docs don't mention any simd
07:11:00 <moon-child> should they? It's an implementation detail
07:11:00 <mrvn> they describe all their other details
07:12:00 <moon-child> anyway I grepped for 'vmovdq' and turned up hits. So they do use simd
07:14:00 <mrvn> $ rgrep vmovdq -l
07:14:00 <mrvn> mpn/x86_64/fastavx/copyd.asm
07:14:00 <mrvn> mpn/x86_64/fastavx/copyi.asm
07:15:00 <mrvn> For some form of "use"
07:18:00 <mrvn> There is quite a bit of use for %xmm but %ymm only shows up in the copy code.
07:21:00 <mrvn> I don't even need the algorithms to use SIMD internally but they should provide verctorized versions of the operations. E.g. add 4 pairs of numbers in parallel using SIMD.
07:22:00 <moon-child> oh, that is annoying to do
07:22:00 <moon-child> better off using adox/adcx
07:25:00 <mrvn> That would let me add 2 pairs in parallel. Wouldn't AVX adding 4 pairs be faster?
07:28:00 <moon-child> maybe. Probably not. Getting overflow is annoying with avx
07:30:00 <mrvn> moon-child: a + b < a
07:30:00 <moon-child> I guess
07:31:00 <mrvn> but yeah, you have to <, shift and then add the carry
07:31:00 <moon-child> you also have to do index management--deal with the case when one of the numbers runs out of limbs before the other
07:31:00 <moon-child> and either gather or manually transpose
07:31:00 <mrvn> Single pair vs. 4 pairs it should be a win. 2 vs. 4 maybe not.
07:32:00 <mrvn> Nah, you would use that with identical sizes numbers. Like when you split a A * B into (A0 A1) * (B0 B1) with power of 2 sizes.
07:35:00 <mrvn> gmp uses adcx/adox for addmul for example.
08:01:00 <mrvn> Am I missing someting or does AArch64 not have a "MUL V0.2D, V1.2D, V2.2D" nor "MUL V0.2D, V1.2S, V2.2S"?
08:32:00 <geist> not sure precisely what you mean there with the .2D stuff and whatnot
08:32:00 <geist> are you talking about particular lanes of a simd vector?
08:33:00 <mrvn> yes. 2 64bit * 64bit
08:33:00 <geist> might be useful to try to see what godbolt emits if you can coax it out. i think there are types for the vector stuff
08:34:00 <geist> the arm64 ASIMD stuff should be a the minimum a superset of the NEON bits, and i think it's pretty regular so i guess it woul dhave what you want?
08:34:00 <geist> but of course the manual is probalby the best bet
08:34:00 <mrvn> https://developer.arm.com/documentation/dui0801/g/A64-SIMD-Vector-Instructions/ADD--vector-?lang=en
08:34:00 <bslsk05> ​developer.arm.com: Documentation – Arm Developer
08:34:00 <mrvn> https://developer.arm.com/documentation/dui0801/g/A64-SIMD-Vector-Instructions/MUL--vector-?lang=en
08:34:00 <bslsk05> ​developer.arm.com: Documentation – Arm Developer
08:35:00 <mrvn> ADD has 2D but MUL does not.
08:35:00 <geist> huh interesting
08:37:00 <geist> it has some multiply+accumulate ones
08:37:00 <geist> maybe you're supposed to use that
08:37:00 <mrvn> Looks like the best one can do is take 4 uint16_t, space them out into 4 uint32_t, multiply and then reassemble.
08:37:00 <geist> that doesn't seem right. surely it has 64bit multiply
08:37:00 <geist> probably just looks like a different form than you expect
08:44:00 <mrvn> MLA has no 2D, but the saturating kinds do have 2D
08:45:00 <mrvn> UMULL, UMULL2 looks like the right one
08:45:00 <mrvn> Did they had to make a new opcode for it when they extended it to 128bit?
08:47:00 <mrvn> Wait, that*s vector * scalar. And 32bit * 32bit = 64bit.
08:48:00 <mrvn> UMULL, UMULL2 (vector) is the right one: https://developer.arm.com/documentation/dui0801/g/A64-SIMD-Vector-Instructions/UMULL--UMULL2--vector-
08:48:00 <bslsk05> ​developer.arm.com: Documentation – Arm Developer
08:49:00 <mrvn> Still no 64bit * 64bit = 64bit with truncation or high/low result.
08:56:00 <mrvn> AArch64 has some strange SIMD opcodes: SQRDMLAH (vector)
08:56:00 <mrvn> Signed Saturating Rounding Doubling Multiply Accumulate returning High Half (vector).
08:59:00 <mrvn> V0 = round(V0 + (V1 * V2) >> 32)
08:59:00 <mrvn> for 4 ints
09:06:00 * mrvn just noticed std::atomic<integral> is defined as two's-complement. No UB with signed overflow.
10:54:00 <mrvn> gcc vectorization kind of sucks: https://godbolt.org/z/djWs53o8G Can't to 4x64bit as two 2x64bit vectors. Clang manages to break it into two vectors.
10:54:00 <bslsk05> ​godbolt.org: Compiler Explorer
10:58:00 <mrvn> 2x64bit vector add produces the same code in gcc and clang, just some reordering. I wonder if it makes any difference.
11:02:00 <mrvn> And what is gcc thinking in the plain "sum" chaning the array index to be 1 based?
11:29:00 <mrvn> cats, always taking over your live^Wnick
11:38:00 <gog> mew
11:43:00 <zid> hi glug
11:43:00 <gog> howdy
12:01:00 * kazinsal headpats gog
12:04:00 * heat goghead pats
13:55:00 * psykose pathead gogs
14:08:00 * Andrew breaks the CTCP ACTIONization of #osdev
14:08:00 <Andrew> This message is not a CTCP!!!
14:10:00 <zid> weird, they changed the lid on soft drink bottles
14:11:00 <zid> and it's incredibly shitty
14:15:00 <mrvn> Do you mean the lids that no longer detach?
16:42:00 <jafarlihi> I've got a thread that prints out to stdout whenever it receives network messages, but this clobbers the unfinished input I'm in the process of typing in my shell. Is there any way I can make it not clobber the input without using something like ncurses?
16:44:00 <zid> add messages to a queue
16:44:00 <zid> and have a single thread responsible for printing things
16:45:00 <jafarlihi> zid: Why that matters? As soon as I printf it will mess up what I'm typing
16:45:00 <GeDaMo> Log to a file?
16:45:00 <jafarlihi> No, I need to see the output while typing
16:45:00 <zid> ah I thought you meant two programs were fighting on the output
16:45:00 <zid> just don't print to the bottom line
16:46:00 <GeDaMo> Two terminals?
16:46:00 <jafarlihi> zid: How?
16:46:00 <zid> this is osdev afterall
16:46:00 <zid> what's your output device?
16:46:00 <jafarlihi> I'm on urxvt Linux
16:46:00 <jafarlihi> Userspace app
16:46:00 <GeDaMo> https://en.wikipedia.org/wiki/ANSI_escape_code#CSI_(Control_Sequence_Introducer)_sequences
16:46:00 <bslsk05> ​en.wikipedia.org: ANSI escape code - Wikipedia
16:46:00 <zid> okay that will just support ansi escapes then yea
16:47:00 <zid> (which is what ncurses is a frontend for)
16:48:00 <jafarlihi> Thanks
16:50:00 <mrvn> press ctrl-l
16:50:00 <jafarlihi> What does it do?
16:52:00 <mrvn> redraw the screen, meanig your input gets printed again.
16:57:00 <heat> mrvn, what's the ascii char for ^L?
16:58:00 <zid> definitely 6
16:58:00 <mrvn> I don't think you can print it.
16:58:00 <heat> it's still an ascii char
16:58:00 <heat> has to be
16:58:00 <zid> 12, btw
16:58:00 <mrvn> It's something the terminal intercepts, like backspace.
16:59:00 <heat> no, it's an ascii char
16:59:00 <heat> all the ctrl+something correspond to ascii chars
16:59:00 <zid> you just add 1 to the caesarian value
16:59:00 <mrvn> heat: that doesn't conflict with what I said
16:59:00 <zid> ^H is backspace, everyone knows that one
16:59:00 <GeDaMo> ^L is formfeed
16:59:00 <zid> ^[ ie escape
17:00:00 <zid> huh mirc actually prints a tab for ^I, weird for a windows program
17:00:00 <heat> why wouldn't tabs work in windows?
17:01:00 <zid> because windows isn't based on terminals and shit
17:01:00 <zid> It's based on the "EDIT" class not readline etc
17:01:00 <heat> irc is for weirdos that like terminals
17:01:00 <zid> ^[ doesn't give me an escape, etc, ^I working is almost certainly mirc implmenting it manually
17:01:00 <heat> also the command line is pretty much ANSI compatible
17:02:00 <heat> maybe even more? idk
17:02:00 <heat> i remember that got a huge fixup in windows 10 like 4 years ago where they improved conhost substancially
18:59:00 <heat> why are we getting irssi ads
18:59:00 <heat> good to know I need to update my package though
19:15:00 <geist> yah saw that too
19:16:00 <geist> also curious what ircv3 features actually are and can they coexist with other ones, etc
19:18:00 <heat> maybe it's like web3 but for irc :)
19:18:00 <geist> yah dunno. also that kinda implies the current irc is v2?
19:19:00 <GeDaMo> Version 3, the search for version 2 :P
19:20:00 <heat> HALF LIFE 3
19:20:00 <geist> some reddit post seems to imply it's basically the same protocol but has some more standardized and advanced authentication and whatnot stuff
20:06:00 <sbalmos> geist: Remember the IRCx protocol extensions that never went anywhere?
20:06:00 <geist> negative
20:07:00 <sbalmos> kind of the same basic idea. pre-SASL for authentication.
20:07:00 <geist> i have always assumed irc was just written in stone. an old prototol like telnet that there's no reason anyone would ever rev
20:07:00 <sbalmos> heh, exactly
20:07:00 <gamozo> Good morning everyone!
20:07:00 <sbalmos> if I can telnet to the port and hand-write a request still, it's good enough for me!
20:10:00 <zid> the protocol's mostly the same but servers have been extending it pretty heavily via modes
20:13:00 <sbalmos> yeah it's like IMAP. The basic protocol's the same. But then half a billion extensions.
20:26:00 <heat> like tcp :)
20:28:00 <geist> right, i guess what i was curious about is if it extended the fundamental protocol at all
20:28:00 <geist> ie, a different low level framing, etc
20:29:00 <geist> there's clearly lots of layers of stuff on top of irc that is more application layer i guess
20:38:00 <zid> irc proto is fairly extensible if you wanted to, but nobody's really bothered
20:39:00 <zid> except sending different numbers at connect time for informational releases
20:39:00 <zid> :from to: msg\r\n isn't exactly going to be hard to extend, but honestly what else is there other than "message" "quit" "join" "part" "mode"
20:41:00 <sortie> Hey proto you say, can do, protobufs
20:42:00 <sortie> wooooo proooootoooooobuffffffs
20:42:00 <gamozo> Hai sortie!
20:42:00 <zid> sortie has googlitis
20:42:00 <gamozo> ^
20:42:00 <sortie> oh no they're onto me
20:42:00 <sortie> quick geist defend me
20:42:00 <zid> from the MIB at google taking you out for leaking your agent status?
20:42:00 <gamozo> What are you gonna do yet, tell us that gmail is by far and away the best email provider?
20:43:00 <geist> PROTOBUFS
20:43:00 <zid> AHH PROTOBUFS
20:43:00 <geist> it is the EL TORITO of bufs
20:43:00 <zid> I HAVE SEEN THE LIGHT
20:43:00 <sortie> BUF THE PROTO
20:43:00 <zid> Glory to protobufs! join us gamozo!
20:43:00 <gamozo> only in #osdev can we find simping for a protocol library
20:43:00 <gamozo> But mom I just wrote a serialization library
20:43:00 <gamozo> I have to justify using it!
20:43:00 <zid> The s in google stands for "simping for protobufs"
20:43:00 <sortie> BEHOLD THE MOVING PROTOS
20:43:00 <gamozo> (it's really hot)
20:43:00 <gamozo> (it's really really hot)
20:44:00 <sortie> We will discontinue one product every hour until you move the protobufs you're supposed to
20:44:00 <gamozo> Doesn't use pointers, only indicies (pooled) and thus the compiler can _merge allocations_
20:44:00 <gamozo> eg, malloc(4) + malloc(4) becomes malloc(8) without having to explicitly write it
20:44:00 <gamozo> hnggg
20:44:00 <gamozo> It's been on my mind for yearssss
20:45:00 <gamozo> (still wip, will probably release it if it works the way I anticipate)
20:45:00 <zid> you invented pool allocators?
20:45:00 <gamozo> A pool allocator without pointers
20:45:00 <zid> You invented quantum computers?
20:45:00 <gamozo> It's more of a type theory (rust) problem than a data structure problem
20:46:00 <zid> ahh so you invented a virtual machine.
20:46:00 <gamozo> Invariant lifetimes allow it to work (safely), I would say it's "inpexressible" in C because it's "too hard to get right"
20:46:00 <gamozo> ya!
20:48:00 <gamozo> Idk if I can replace protobufs for my usecases with it? Maybe? idk
20:48:00 <gamozo> we'll see if it turns out to be slow as shit lol
20:49:00 <geist> get more computers sheesh
20:49:00 <geist> it's the google way
20:49:00 <zid> Invent a new, crappier, language
20:49:00 <zid> then just 'fix it all with protobufs'
20:49:00 <zid> then discontinue support
20:50:00 <geist> build a language that only does protobyufs
20:50:00 <zid> then bring it back under a different name but this time paid for
20:50:00 <zid> then discontinue it again
20:50:00 <zid> THAT is the google way.
20:50:00 <gamozo> "it scales"
20:50:00 <geist> oh yeah soundslike a true googler
20:50:00 <geist> are you sure you dont work here?
20:50:00 <gamozo> "Have you considered taking on people under you?"
20:51:00 <gamozo> "Can you fill this slot on the team"
20:51:00 <gamozo> "We can't lose this opening"
20:51:00 <geist> these are trigger phrases. the fnord of google
20:51:00 <gamozo> "It's okay, if they're bad we can move them to another team"
20:51:00 <gamozo> > Doesn't move them to another team
20:53:00 <zid> You can take me on under you if you like, I'll do 2 days a week for $30k
20:53:00 <zid> no C++, non-negotiable
20:54:00 <gamozo> Tbh, that's a fairly good deal for good quality code lol
20:54:00 <gamozo> 2 days a week is a pretty good level of output :D
20:54:00 <zid> That's 2 work days, 9-5 with a 3 hour lunch
20:54:00 <gamozo> That's a good week for me
20:54:00 <gamozo> lol
20:55:00 <zid> also I get my statutory 21 days paid vacation still
20:55:00 <gamozo> take as many as you want, just get the work done
20:55:00 <zid> that's not what holiday means, americans man
20:56:00 <sortie> 220 files changed, 23390 insertions(+), 2405 deletions(-) ← One of those days
20:56:00 <zid> https://www.tastefullyoffensive.com/wp-content/uploads/2021/05/americans-work-ethic-other-countries-1.jpg
20:56:00 <geist> heh word
20:57:00 <gamozo> Tbh, it's pretty bad
20:58:00 <geist> yah and i even internalize that. i have a tough time taking vacation if i have pending things to do
20:58:00 <sortie> 30 paid vacation days + holidays
20:58:00 <geist> so i always delay vacation until i get something done, and then i just have more to do
20:58:00 <geist> even though i have a crapton of vacation hours queued up
20:58:00 <gamozo> It's rough having to do weekly/monthly statuses work. Like, maybe I'm just defective. But I get like 3-4 months of hyper productivity a year, followed by 8 months of just wanting to chill and work on side projects
20:59:00 <gamozo> and I'm way more productive doing bursts, rather than trying to "pretend" to have consistent output year-round every week the same level of producitivty
20:59:00 <zid> better than my 3-4 days and 11 months
20:59:00 <zid> check my git history
20:59:00 <geist> gamozo: yah agreed
20:59:00 <geist> and the 8 months i just feel bad about not doing as good as the productive ones
20:59:00 <gamozo> I recently formed a business with my friend, because I cannot do effective technical work at $BIGCORP
20:59:00 <geist> because of the internalized shame at it
20:59:00 <gamozo> and technical work is what I _love_
21:00:00 <gamozo> It's so hypeee
21:00:00 <zid> can I help, I will work for $30k, 2 da...
21:00:00 <gamozo> :D
21:00:00 <geist> and companies like $BIGCORP force you to deal with it by having this over the table performance process that forces you to write everything down, etc
21:02:00 <gamozo> Yeah, I think it works great for people who need the guidance, but for people who don't it really hurts them I think
21:02:00 <zid> does your job have the cool feature american jobs have
21:02:00 <zid> where if you do your job really well
21:02:00 <zid> you get to do other people's jobs as well
21:02:00 <gamozo> sadly (from what I've seen, could be my specific teams) beyond senior you're forced to manage, and senior is often 4-6 year of exp. So the "technical" path after just a few years of technical work :(
21:03:00 <geist> yah not necessarily forced i think, but at least forced to 'architect' or whatnot
21:03:00 <gamozo> Like, I like architecting, but I can't even get that work
21:03:00 <geist> ie, less hands on technical work and more writing things down technical work
21:03:00 <geist> and i think that's a different skill set
21:03:00 <gamozo> I've offered that but I don't have leadership experience so they won't let me do it (without proving myself). Which don't get me wrong, that's fair
21:04:00 <gamozo> but I don't want to do a leadership role to prove myself at senior pay
21:04:00 <gamozo> maybe I'm greedy, but I can do that grind doing consulting with direct rewards for my technical output
21:04:00 <geist> that's why i fiddle with osdev or hobby stuff on the side. gives me the satisfaction of technical stuff without any of the un-fun BS parts of it
21:05:00 <gamozo> The problem is my technical experience is maybe 15-20+ years, but my leadership experience is 0, so to $BIGCORP, max I'm senior (which is usually just 6 years technical exp or some shit)
21:05:00 <gamozo> feels bad cause I'm pretty sure I could do the leading, if I knoew I wasn't doing it for insanely cheap
21:06:00 <gamozo> I love teaching and technical leadership, but I do that on my free time with my outside-of-work projects, and they don't trust that experience I guess? Idk
21:06:00 <gamozo> which is fair, it's a bit "easier" than some of the stuff I'd deal with at a company
21:06:00 <gamozo> /rant
21:06:00 <gamozo> (this maybe has been on my mind this past year as like a big hurdle)
21:07:00 <geist> hrm, got a mysterious shipment package notification from china that i dont know what it goes to
21:07:00 <geist> but then i have a set of preorders for various boards out that i've half forgotten about, so it could be any one of them
21:07:00 <zid> ooh christmas presents
21:08:00 <gamozo> Surprise!
21:08:00 <zid> past you was kind to future you
21:08:00 <zid> past me is an asshole
21:08:00 <zid> he never buys me toys
21:08:00 <geist> i think it might be one of the riscv boards i ordered a while back
21:08:00 <zid> oh at least you got a bad toy
21:08:00 <zid> now I feel less envious
21:08:00 <geist> i ordered one of the rockchip router boards i linked the other day but that wsas explicitly DHL based
21:08:00 <geist> this is china post
21:09:00 <gamozo> arm64?
21:09:00 <geist> the router thing? yeah
21:09:00 <geist> one of these: https://www.friendlyelec.com/index.php?route=product/product&product_id=287
21:09:00 <bslsk05> ​www.friendlyelec.com: NanoPi R5S
21:09:00 <geist> did some digging around and the docs are basically fully available for the rockchip in it, and looks pretty solid
21:11:00 <geist> being actually available it seems like a nice product if you're okay with it being headless. there's some other board with the same soc that has hdmi and whatnot brought out
21:14:00 <geist> hmm, it's not the risc-v board. thats *still* on preorder (Vision-Five)
21:22:00 <zid> what's one of those
21:28:00 <j`ey> geist: that one you linked shows it having hdmi
21:44:00 <heat> you idiots use protobufs? ha
21:44:00 <heat> we use capnproto, literally protobufs v2
21:44:00 <heat> built by the same guy
21:45:00 <zid> I use structs
21:45:00 <zid> built by donald knuth
21:46:00 <heat> but that does not scale
21:46:00 <heat> checkmate zid
21:47:00 <zid> struct buf { int scalefactor; };
21:47:00 <zid> just increment that, np
22:07:00 <geist> heat: yeah i remember hanging out at Kenton's house a few times (he had this epic gaming house at the time) about when he was publishing captnproto
22:08:00 <geist> ah he works at cloudflare now. you'll probably meet him
22:08:00 <heat> oh you know him?
22:08:00 <heat> he's on my team :)
22:08:00 <geist> yah a bit, though i only hung out at his house a few times with a bunch of other people
22:08:00 <geist> so i dunno if he remembers me well
22:09:00 <geist> he had this epic LAN house
22:09:00 <geist> https://kentonshouse.com/
22:09:00 <bslsk05> ​kentonshouse.com: Kenton's House
22:09:00 <geist> he had this i think biweekly gaming day at his house which was really fun
22:10:00 <heat> woah :D
22:11:00 <heat> did he just invite people from the office?
22:13:00 <geist> or friends of friends, etc. it was more like you signed up no his page and brought food and beer
22:13:00 <geist> i think at the time he had just left or recently left google, and it was long before i worked there
22:14:00 <heat> must've been in 2013
22:14:00 <geist> yah maybe i was at google at the time
22:15:00 <geist> i remember teh tech design was kinda interesting, i tink he had 12 statinos around the house, mostly hidden away in wall panels that you pulled down. the desktops themselves were rack mounted all n one closet
22:15:00 <geist> he had some thing where each had a copy of windows that was running off a central nas box, each was iscsi mounted the root drive
22:16:00 <geist> and operating on some sort of snapshot, so at the end of the day when you shut them down it reset back to scratch
22:16:00 <geist> i dunno precisely how he ran the hdmi and usb and whatnot across the house, but there's a whole tech page on that site
22:16:00 <geist> all told the whole thing wans't *that* expensive. like $20k or something
22:16:00 <geist> all considering
22:17:00 <gamozo> This is cool!
22:18:00 <geist> iirc the iscsi snapshot stuff was pretty cheezy and effective. the iscsi machine was serving 12 LVM partitions off a disk, and each LVM partition was set up at the start of the day as a COW snapshot of a golden one
22:18:00 <heat> https://kentonsprojects.blogspot.com/2011/12/lan-party-house-technical-design-and.html
22:18:00 <bslsk05> ​kentonsprojects.blogspot.com: Kenton's Weekend Projects: LAN-party house: Technical design and FAQ
22:19:00 <heat> geist: the fact you can recall technical details of this lan party thing you went to 10 years ago is pretty impressive lol
22:21:00 <geist> heh i remember being pretty interested in it
22:21:00 <geist> also was surprised to learn that windows has pretty decent built in boot-from-iscsi support
22:23:00 <heat> the google+ references on his blog are mindnumbingly funny to me
22:24:00 <geist> ah yes.
22:24:00 <zid> they're very cuil
22:24:00 <geist> i remeber actually liking g+ once i culled it to be just technical blogs and whatnot. didn't get popular enough to turn into a social media cesspit
22:25:00 <heat> maybe one day I'll ask him "hey Kenton remember this operating systems weirdo from Google that kept blabbering on about iSCSI while everyone wanted to play vidya games" :P
22:25:00 <geist> hah no there were lots of techies there
22:25:00 <geist> this is silicon valley
22:26:00 <zid> "guy didn't even know what an 8 pool was"
22:26:00 <mjg_> 3 rax
22:26:00 <heat> I would assume operating systems people are a small minority no?
22:26:00 <zid> I tried playing sc2 recently, having not played it for like.. a decade
22:26:00 <mjg_> funny you say that
22:26:00 <zid> it felt *so* weird, like I was playing it through a mail slot using someone else's arms
22:27:00 <mjg_> i was almos flirting with the idea this week, then i learned that frostgiant (after blizzard exodus) is going to have a beta new rts soon
22:27:00 <mjg_> and people are going to switch, mostl ikely
22:28:00 <mjg_> even so, apparently the population count is at about 200k right now
22:28:00 <zid> I got to diamond when there was only diamond, and then very bottom of master when they added that
22:28:00 <mjg_> i though the game would be way more dead right now
22:28:00 <geist> which game would be dead?
22:28:00 <zid> sc2
22:29:00 <zid> https://kentonshouse.com/photo3.jpeg This game
22:29:00 <mjg_> mandatory https://www.youtube.com/watch?v=SAo3QwnPruc
22:29:00 <bslsk05> ​'The hands of God - Flash APM/First Person video' by KhaldorTV (00:09:43)
22:29:00 <heat> it looks like someone's got a terminal open back there wtf
22:30:00 <zid> how dare they
22:30:00 <geist> so what i remember playing in that room was some star trek like simulator thing whee everyone had a different console
22:30:00 <heat> also looks like ubuntu
22:30:00 <heat> the real crime
22:30:00 <geist> it was really fun. each person controlled one part of the station, and the captain sat in the middle and had the big screen
22:30:00 <geist> but couldn't really directly control anything, but could tell other people what to do
22:35:00 <geist> ah i think it was https://store.steampowered.com/app/247350/Artemis_Spaceship_Bridge_Simulator/
22:35:00 <bslsk05> ​store.steampowered.com: Artemis Spaceship Bridge Simulator on Steam
22:36:00 <mjg_> did you set shield to maximum
22:36:00 <mjg_> or is this the wrong franchise
22:37:00 <geist> iirc that was part of the challenge. one of the stations was engineering so i think you had to control power distribution
22:45:00 <gamozo> attic?
22:45:00 <gamozo> wow wrong window
22:48:00 <zid> gamozo: where do you keep your bound captives?
22:48:00 <zid> wow wrong window
22:48:00 <gamozo> NO ITS LUNCH PLANS OKAY GOSH
22:48:00 <zid> protesting pretttty hard there
22:48:00 <heat> lunch in the attic? sounds sus
22:54:00 <gamozo> :3
22:54:00 <gamozo> It's a cool little restaurant over a waterfall
22:54:00 <zid> It always is, if it's called yon attic
22:54:00 <gamozo> Are you implying I do waterfall?
22:57:00 <zid> It's always a little resturant
22:57:00 <zid> never something fun
22:57:00 <zid> like a lava moat
22:57:00 <heat> size doesn't matter
22:57:00 <heat> little restaurants can be fun too
23:00:00 <zid> That's what someone with a little resturant would say
23:00:00 <heat> hey it's an average size restaurant
23:00:00 <zid> idk, you seem to be compensating
23:23:00 <gog> you want lava moat
23:23:00 <gog> i'll get you lava moat
23:28:00 <kingoffrance> </cave johnson gog>
23:42:00 <zid> can we confine it to just my attic
23:43:00 <zid> and have it roll out into the garden in the summers when it's hjot
23:45:00 <gog> sure
23:52:00 <zid> are the refill cartridges affordable? That's how they get you