02:14 |
crtdaydreams |
probably why, home pc nat not working |
02:15 |
crtdaydreams |
likely on our end if ascii can talk with dulapbot |
| |
~ 9 hours 6 minutes ~ |
11:21 |
awt |
cgra et al.: http://paste.deedbot.org/?id=UXKZ << patch to send ac only on launch/new key/address change for review |
| |
↖ |
11:28 |
phf |
http://logs.bitdash.io/pest/2023-02-16#1023375 << i'd check the antecedants to the missing packets, because if they are from my client, then they are lost FOREVER |
| |
↖ |
11:28 |
bitbot |
Logged on 2023-02-16 06:29:41 cgra[jonsykkel|awt|signpost]: when looking, found out that there were ~800 unresolved chain hashes in my db. after some manual getdata flood at my peers i ended up with ~10k (doubling the total) more messages and unresolved chain hashes are now down to 12. these 12 won't budge |
11:29 |
phf |
i'm pretty sure i've also sent at least one or two malformed packets, where the netchain or self-chain or both would've been some totally random sequence |
11:31 |
phf |
it's either that, or having packet loss before i have a working getdata: packet A is lost, packet B goes out, counterparty getdata's A, but the client doesn't respond. |
11:31 |
phf |
*before i had |
11:33 |
phf |
http://logs.bitdash.io/pest/2023-02-16#1023376 << that's some strange dulapbot bug, because the messages in question are not in pest logs |
11:33 |
bitbot |
Logged on 2023-02-16 06:32:31 dulapbot[jonsykkel|awt|deedbot]: (trilema) 2016-09-16 phf: in case anyone thinks that commercial lisp vendors are any good, i'm getting completely random failed to allocate errors on lispworks |
11:33 |
phf |
or are they? they are not in bitdash history anyway |
11:34 |
phf |
but for the rescind the point above, i figured out the random to allocate errors and it was my own doing |
11:34 |
phf |
*but for the record |
11:35 |
phf |
fok, i need my morning tea |
11:40 |
asciilifeform |
phf: it's likely a bug in the dulapbot+blatta ensemble, i.e. the former fires when sees a replay just the same as if realtime msg. |
11:40 |
asciilifeform |
evidently saw buncha replays earlier. |
11:41 |
asciilifeform |
for that matter, this is all over the dulapbot #p log, elsewhere. |
11:43 |
asciilifeform |
http://logs.bitdash.io/pest/2023-02-18#1023634 << was asciilifeform's 1st guess, but wonder what accounts for the others |
11:43 |
bitbot |
Logged on 2023-02-18 11:28:46 phf[awt|deedbot]: http://logs.bitdash.io/pest/2023-02-16#1023375 << i'd check the antecedants to the missing packets, because if they are from my client, then they are lost FOREVER |
11:43 |
bitbot |
Logged on 2023-02-16 15:56:48 asciilifeform[jonsykkel]: cgra: some of these are likely phf's proto station with the hand-cranked packet shooter. but the rest -- mystery |
11:44 |
phf |
a surprisingly small number (i got to that point in log) |
11:45 |
asciilifeform |
( for reference, subj ) |
11:45 |
bitbot |
Logged on 2023-02-16 15:55:04 cgra[jonsykkel|awt|signpost]: http://logs.bitdash.io/pest/2023-02-16#1023381 << here, and somehow it's 13 now |
11:48 |
phf |
for how small that number is, almost a shame that it's there at all :> |
11:48 |
* |
asciilifeform also expected there'd be moar of'em |
11:49 |
phf |
http://logs.bitdash.io/pest/2023-02-17#1023443 << afaik thunderbolt 4 machines https://support.apple.com/en-us/HT201736 support up to 4 external monitors |
11:49 |
bitbot |
Logged on 2023-02-17 11:45:51 asciilifeform[4]: phf ( or any other folx who operate crapple irons ) -- know whether crapple sells a box that can drive 3 'displayport' outputs ? |
11:50 |
phf |
but then as far as i understand question becomes can the internal gpu drive the total surface space of those monitors? i'm not quite sure how it works |
11:50 |
asciilifeform |
phf: their docs re the 'mini' brick (newest 1, with such ports) seem to suggest 'drives 3, but 1 must be via hdmi' |
11:51 |
asciilifeform |
so in fact loox like the only crapples that can pump out 3 displayports are 'studio', 'pro' lappy, and 'cheese grater' |
11:51 |
asciilifeform |
( of the arm boxen, that is ) |
11:52 |
phf |
if i were building a mac powerstation, and lets say i used one of those "sealed boxes" as a base, like macbook or mac mini, i'd go the egpu route |
11:53 |
asciilifeform |
phf: per crapple docs, egpu only worx on intel crapples |
11:53 |
phf |
well, you know about it more than i do then |
11:54 |
asciilifeform |
was trying to find out before buying paperweight for its weight in gold, lol |
11:55 |
* |
asciilifeform already bought, lol, 1 paperweight |
11:55 |
bitbot |
Logged on 2023-02-17 12:56:37 asciilifeform[jonsykkel|awt|deedbot]: has also been carrying on a crapple port of gui pestron proto (for use on lappies) but developing on lappy has proved physically unpleasant |
11:56 |
phf |
i've just assembled an PC COMPUTER (pronounced https://www.youtube.com/watch?v=Q77YBmtd2Rw&t=25s) and marveled at how fucking insanely fast it is |
11:57 |
phf |
as our mutual bolix friend said "oh you just noticed a difference between a 45W gaytop and a 150w TDP desktop chip??" |
11:57 |
asciilifeform |
phf: i've a '$maxint' 'threadripper' in there, and the m2 el cheapo in fact beats the shit outta it |
11:57 |
bitbot |
Logged on 2022-11-23 13:38:50 asciilifeform[5]: phf: for comparison: 2m 49s on 'threadripper' 3960X. |
11:57 |
bitbot |
Logged on 2023-02-17 13:04:01 asciilifeform[4]: ( tho, interestingly, that thing is in fact the fastest (single-threaded, at any rate) box in asciilifeform's torture room, lol |
11:58 |
* |
asciilifeform was convinced there's some kinda fraud in published benchmarks, ended up in fact buying the box to see for himself. wasn't a fraud. |
11:59 |
asciilifeform |
x86 suxx, 'news at 11'.. |
12:00 |
asciilifeform |
even compile (gnat) is ~5x faster on the 'm2' vs 'threadripper' |
12:01 |
phf |
asciilifeform, i should probably run some benchmarks.. e.g. are you comparing "like for like" |
12:01 |
asciilifeform |
granted, it's a 2020 threadripper. |
12:02 |
phf |
well, no, what's you memory freq, are you compiling it off scsi ssd or m.2 |
12:03 |
phf |
e.g. your threadrippre is probably ecc, while m1 is not, that ought to add overhead, particularly on such "tight loop" ops |
12:16 |
phf |
like if you're getting 5x compile speed differences Something Else Is Going on, for sure. i can buy that m1 is faster than equivalent or previous small device cpus by a margin |
12:17 |
asciilifeform |
phf: 'samsung 970 pro' nvme ssd in the 'threadripper'. sadly dun recall the precise memory bw |
12:17 |
* |
asciilifeform will dig when gets a chance |
12:18 |
phf |
i have openbsd running on my m1 (which was the original reason for even buying it, since long forgotten |
12:18 |
asciilifeform |
phf: the benchmark in question was (see thrd) ffa, which strongly suspect fits in l1 on both irons. so prolly aint the memory bus. but hard to say for certain. |
12:19 |
phf |
) so i might be able to setup a like for like benchmark |
12:19 |
asciilifeform |
phf: asciilifeform considering to buy the 'studio' brick, and bsd etc. but not knows whether the gpu has been reversed |
12:20 |
phf |
no, def not to the point of running it with multiple HDMI monitors or anything like that |
12:20 |
asciilifeform |
recipe for that benchmark for ref. |
12:20 |
bitbot |
Logged on 2022-11-23 14:15:00 asciilifeform[6]: ftr, how to run: 1) throw out the flags (e.g. '-static') that dunwork on crapple gnat; 2) in limits.ads, up the max tape from 1MB to 12MB 3) up the stack to max crapple permits: ulimit -s 65520 3) : |
12:20 |
* |
asciilifeform bbl -- teatime |
12:25 |
phf |
oh, i thought you were saying ~gnat compiles~ 5x faster |
| |
↖ |
12:26 |
phf |
http://logs.bitdash.io/pest/2022-11-23#1016815 << that particular tape has not been published afaiu |
| |
↖ |
12:26 |
bitbot |
Logged on 2022-11-23 14:15:08 asciilifeform[jonsykkel|deedbot]: time cat tapes/10k_shots_1024bit_ffa_slid_rnd.tape | ./ffacalc/bin/peh 1024 32 10660000 0 |
12:33 |
phf |
http://logs.bitdash.io/pest/2023-02-17#1023484 << "hacker" "culture" |
12:33 |
bitbot |
Logged on 2023-02-17 13:03:53 signpost: in unrelated news, phf is right. half of "AI" is trans weebs. |
12:35 |
phf |
http://logs.bitdash.io/pest/2023-02-17#1023489 << i'm going through this article https://jaykmody.com/blog/gpt-from-scratch/ at the moment |
12:35 |
bitbot |
Logged on 2023-02-17 13:05:24 signpost: booting up an opensource turd gpt |
12:35 |
phf |
unfortunate side-effect of PC COMPUTER is that it runs cyberpunk 2077 as well as it does numpy |
12:39 |
phf |
right above quoted line there was a conversation about apple computer leasing and tivoization and all that, which ties with the gpt https://archive.is/PUShr |
12:40 |
phf |
"OpenAI Proposes Government Restrict AI Chips to Prevent Propaganda Explosion" (i haven't read the article, but it appears to summarize and link to relevant openai paper) |
| |
↖ |
12:42 |
phf |
funny thing when we had the previous conversation about chatgpt, i was going to make a cyberpunk themed joke, "oh yeah they are going to classify high-performance gpus as munitions so that you can't make your chatgpt say bad words at home" |
12:43 |
phf |
i since found out that high-performance gpus already have export restrictions (kind of obvious, but i didn't think about it), and of course openai people being what they are they are proposing such measures in earnest |
| |
↖ |
12:44 |
phf |
http://logs.bitdash.io/pest/2023-02-17#1023481 |
12:44 |
bitbot |
Logged on 2023-02-17 13:02:18 billymg: asciilifeform: you vill own nothing and you vill be happy |
12:51 |
billymg |
http://logs.bitdash.io/pest/2023-02-18#1023700 << recalls 90s apple ad for G4 desktop |
12:51 |
bitbot |
Logged on 2023-02-18 12:40:26 phf[awt|deedbot]: "OpenAI Proposes Government Restrict AI Chips to Prevent Propaganda Explosion" (i haven't read the article, but it appears to summarize and link to relevant openai paper) |
12:55 |
billymg |
nvidia also supposedly cancelling plans for their 48GB consumer card, can't help but think it's related |
12:55 |
phf |
that ad is relevant, but man what a lame ad. just stylistically and thematically |
12:59 |
asciilifeform |
http://logs.bitdash.io/pest/2023-02-18#1023692 << it was in a tarball in ch14 |
12:59 |
bitbot |
Logged on 2023-02-18 12:26:46 phf[awt|deedbot]: http://logs.bitdash.io/pest/2022-11-23#1016815 << that particular tape has not been published afaiu |
13:00 |
asciilifeform |
http://logs.bitdash.io/pest/2023-02-18#1023691 << ftr was 'naked eye' observation, there, asciilifeform not timed the compile. |
13:00 |
bitbot |
Logged on 2023-02-18 12:25:08 phf[awt|deedbot]: oh, i thought you were saying ~gnat compiles~ 5x faster |
13:01 |
phf |
in other news, asciilifeform has switched jobs and now is working for apple corporation! |
13:01 |
asciilifeform |
lol satan preserve |
13:04 |
asciilifeform |
http://logs.bitdash.io/pest/2023-02-18#1023702 << the very phrase 'export restrictions' invokes '80s nostalgia, y'know, back when serious irons were actually stamped out in the reich proper |
13:04 |
bitbot |
Logged on 2023-02-18 12:43:56 phf[awt|deedbot]: i since found out that high-performance gpus already have export restrictions (kind of obvious, but i didn't think about it), and of course openai people being what they are they are proposing such measures in earnest |
13:04 |
asciilifeform |
'verboten to ship a microvax to su!' etc |
13:06 |
phf |
the "unlicensed site bugfix" folklore stories |
13:06 |
asciilifeform |
kremvax(tm)(r) ! |
13:08 |
phf |
i feel like at some i posted homepage of somebody from kurchatov talking about ins and outs of running custom whatever it was pdp at a closed research facility |
13:09 |
asciilifeform |
re gptism -- asciilifeform played, like errybody else, with the public demo, and still can't muster much moar excitement than had for 'eliza' in '90s. gpttron does a kinda 90% emulation of the segment of the net people asciilifeform not cares to interact with, lolyes. but this was good for an evening's lul, at most |
13:10 |
asciilifeform |
phf: not recall the pdp link, but possibly. (and if one believes rumours, to this day they're baking bk0010-style micropdp dies in zelenograd, for the 'land that time forgot' eternal 1980s reenactments) |
13:11 |
* |
asciilifeform unearthed a 'mk-85' pocket calculator at one pt, with that same chipset. runs on a watch battery installed in 1990 (seal still intact, lol), a++ |
13:12 |
phf |
"you think you need more than pdp in your self guided rocket control? you have desease of american capitalist!" |
13:12 |
asciilifeform |
^ |
13:13 |
asciilifeform |
fwiw reich's 'minute man 3' reportedly still pdp11. |
13:13 |
phf |
my impression is further up the critical infrastructure chain you go the more "still X" you get |
13:14 |
asciilifeform |
( or at least claimed. likely if they confessed that it runs win10 , errybody else will pull triggers , lol ) |
13:14 |
asciilifeform |
phf: indeed |
13:14 |
* |
asciilifeform recalls old #t thrd re pdp11 install base in nuke power industry |
13:15 |
phf |
my favorite one is nyc subway running on os/2, which, and i mean the story, i find to be aesthetically pleasing |
13:16 |
* |
asciilifeform not so many yrs ago, still had a working os2 box. 'galciv' ! |
13:17 |
asciilifeform |
iirc nyc subway until mid-90s still ran on the original mech relays. |
13:17 |
asciilifeform |
and until '80 or so, powered by orig. steam dynamo. |
13:17 |
asciilifeform |
'if it worx, dunfixit' |
13:20 |
phf |
if it worx, have a vendor rebuild it in node.js for $xmil |
13:20 |
asciilifeform |
inevitable |
13:22 |
* |
asciilifeform recalls '1st swallows of spring' there, e.g. boeing's cppization and luloverflow bugola etc |
13:23 |
asciilifeform |
the folx who not grew up on lol.js etc and can actually program 'on watch battery installed in 1990' dun live 4evah. inescapably will be replaced with what's avail. to replace'em with. |
13:24 |
phf |
i don't think that it's just "kids these days". as some people like to point out "there's more freaks online now than used to be, but the noise is higher" |
13:24 |
asciilifeform |
phf: is moar of a 'you are what you eat' problem, unfortunately |
13:25 |
phf |
e.g. poc||gtfo zoomers program on watch battery for fun, then go to design python backends for adtech |
13:27 |
asciilifeform |
cultural demise. recall thrds re 'they then bring adtech habits to emacs' etc |
13:28 |
asciilifeform |
'soar with the pigs by day and wallow with the eagles at night' turns out not worx so well in practice. |
13:29 |
phf |
i think it's just the sheer number of pigs |
13:30 |
phf |
and the application of computing where it need not be |
13:30 |
asciilifeform |
oblig |
13:31 |
asciilifeform |
grr wat |
13:31 |
asciilifeform |
index off, apparently |
13:31 |
asciilifeform |
let's try that again |
13:32 |
asciilifeform |
aand nao it's in buffer limbo, lol |
13:32 |
phf |
yeah i mean smarttv, why. but there's a hacker somewhere there in the mix trying to get a bunch of chatgpts to produce code that works and hold it all together |
13:32 |
dulapbot |
(asciilifeform) 2022-04-22 asciilifeform: ( meanwhile you prolly can't buy a flashlight w/out a computer. and crashes. Because Reasons ) |
13:32 |
dulapbot |
(asciilifeform) 2020-05-10 asciilifeform: i've found flashlights w/ 8bit micros in'em. |
13:33 |
asciilifeform |
phf: from 'heighten the contradictions'(tm)(r) pov -- a++. wainot, let's crank out 'indian coad' with subtle bugola on erry 3rd line at speed of gpu instead of speed of indians, wainot |
13:38 |
phf |
well, the point i want to make but keep not making, is that you have diffusion of talent, combined with misallocation of talent. in the 70s everyone who's anyone was at a handful of "unix rooms", mit ai lib, xerox, bell labs. these day neither unix room nor the kind of tasks it was set to solve are economically viable. |
13:41 |
asciilifeform |
phf: (iirc beaten to a painful death in ye olde #t bug) see also pike 2k, where evident that the 'egyptian stagnation' already underway then. |
13:41 |
asciilifeform |
*#t but.. |
13:41 |
phf |
yah |
13:42 |
phf |
subj has been beaten to death, also old man yell at clowd |
13:44 |
asciilifeform |
fwiw it prolly is still possible to learn 'on watch battery' from old b00kz & artifacts , and then go on even to program for dough w/out ever touching 'adtech & loljs' , just as was possible when asciilifeform started. but the door to that world aint getting any wider or easier to open, quite opposite |
13:45 |
* |
asciilifeform fwiw never sat in the 'unix rooms', set foot in a 'mit', etc |
13:46 |
asciilifeform |
or take e.g. a fella asciilifeform once taught & worked with, who was booted outta high school, drove truck by day and learned to reverse at night on 'warcraft 2' adhoc net, still going in 2010s, lol |
13:46 |
asciilifeform |
walked in with 0day for subj, lol, which got him hired |
13:48 |
asciilifeform |
such things , suspect, remain possible, but the # of cracks in the pavement you gotta fall through is climbing. |
13:48 |
phf |
i know some ru hackerspace zoomers, who are talented hackers, they are all working for different paid in shitcoin cryptoscam builders |
13:50 |
asciilifeform |
hey, 1 kind of salt mine where harvesting printola from idjits is moar or less same as anuther |
13:50 |
asciilifeform |
the sad part is that no 'state of art' is advanced. but this -- same nao as 20y ago |
13:52 |
* |
asciilifeform must bbl |