Show Idle (>14 d.) Chans

← 2017-10-04 | 2017-10-06 →
mircea_popescu: in other great news, looks like the politruks got together, decided they really only have the office space for one of the two, and axed twitter.
mircea_popescu: moving towards "advertising agency" in desperation is perhaps the lulziest fiat conceit yet.
mircea_popescu: not only for the direct reason, but for the implicit hope that maybe... wikipedia follows suit ?
mircea_popescu: all the fucktards going around wikipediaing for the past decade won't chase me around so i step on their worthless bodies instead of the pavement, of course.
BingoBoingo: Hey, walmart is moving to online strategy of "Amazon 3rd party reseller market" but shittier website
mircea_popescu: nor will they ever raise the acumen to confront the plain reality of how they ~should~
mircea_popescu: but i'll lol, which is comparatively the larger part anyway.
BingoBoingo: I wonder how pissed average USian would be if they learned large chunks of the world gets 6%+ economic growth sustained over decades
a111: Logged on 2017-10-05 04:03 mircea_popescu: in other great news, looks like the politruks got together, decided they really only have the office space for one of the two, and axed twitter.
trinque: << "Two black lawmakers in the U.S. Congress" something or other
asciilifeform: this is pretty great
mircea_popescu: !!withdraw 0.0120199 14PywEwC84oBNn4fZiFJssjDPQTokgdGE7
mircea_popescu: !!v A3C339306287A0EF5DB3FF8AAE9B8B025273132697C1BCB814C3EFC1AE30E984
mircea_popescu: trinque around ? :D
trinque: this urgent, or can wait til tomorrow?
midnightmagic: cool deedbot functionality.
mircea_popescu: not really urgent if inconvenient
mircea_popescu: asciilifeform whatever, "investors displeased with key figures" if you prefer the in-universe formulation
asciilifeform: also great, can't wait to buy ticker to watch the cement poured
asciilifeform: *ticket
mircea_popescu: nah, likely will unwind the way of ~any other scam
mircea_popescu: "bought by verizon"
trinque: mircea_popescu: get it out for ya tomorrow with a fresh brain; about to retire for the night
mircea_popescu: "Venture capital isn’t an asset class, it’s a system to transfer wealth from state pension funds into the pockets of general partners. Some of the wealth moves into the pockets of kids working on new technology. This part is important, because that technology can have a moonshot chance of improving the world. Someone’s gotta fund moonshots, and it’s not gonna be banks. Venture capital, then, is a charitable service. W
mircea_popescu: hy not advertise honestly? Nonprofits are allowed a 3% management fee so VC firms can still afford disgusting salaries and daily crudité delivery. Maybe that decreases the likelihood that endowments and pension funds allocate assets to Silicon Valley, but LPs allocate less than 0.5% of their funds to this subset of private equity anyway. Venture capital is already a tax writeoff." << check out who gets how the whole pile of
mircea_popescu: irrelevant noise actually works.
BingoBoingo: Is that Elaine?
mircea_popescu: a yeah, what happened to her ?
BingoBoingo: I'm not sure
mircea_popescu: in other lulz, wall funded in congress ?
BingoBoingo: Not yet, just passed a house committee
mircea_popescu: ah right.
mircea_popescu: o look, bitcoin crash under 0.085 nao.
BingoBoingo: lol, a lot of people would be surprised how bouncy concrete is
mircea_popescu: and a lot of redditors would be surprised at how bouncy tits are.
mircea_popescu: experience is a valuable teacher absent for most experts.
BingoBoingo: Bouncy's one of those weird qualities.
mircea_popescu: aaand in other "after graduation i plan to spend a year in africa doing charity work" news,
mod6: mornin'
mod6: hola
mircea_popescu: hey mod6
mircea_popescu: aaand in other "global warming" lulz, costa rica has been beset by an... anti-tropical storm, with temperatures 10 degrees below period averages and winds in the 50kph+ range since yest.
mircea_popescu: apparently this is now possible.
mod6: heh
mod6: pretty windy
shinohai: mircea_popescu should hold kite-flying contest amongst the slaves. Winner is girl who can keep kite aloft for 2+ minutes using only erect nipples.
mircea_popescu: the avian biodiversidad is pretty fucking sad looking, i'll tell you that.
mircea_popescu: these pair of eagles came to say good morning erry day, nowehre to be seen this morn.
mod6: hunkered down huh
mircea_popescu: i don't expect they can fly in this. eagle isn't a swallow.
mircea_popescu: and in other lulz : (poor dooglus ended up paying a 4k btc to a bunch of lazy retartds pretending to be strippers, two years after.)
mircea_popescu: an honest man, he is ; actually functional in any proper sense, they aren't.
shinohai: If I made that offer on some South-American camslut site, it wouldn't last a week.
mircea_popescu: there's half a billion of them, and what, one a year manages to find her way to a hundy ?
mircea_popescu: i don't think it can be put into words, how much of the whole "poor people need help" the pay-for-your-tits disabused in my mind.
mircea_popescu: they need help like they need designer bags.
asciilifeform: designer body bags
asciilifeform: in other noose, asciilifeform derived a simple constant-time variant of barrett. short enough to put in the l0gz:
asciilifeform: algo computes r := x mod m.
asciilifeform: or rather, x0 mod m, x1 mod m, ... ( as happens in modexp )
asciilifeform: let, for concreteness, x's are 8192 bits wide ( as they are in the 4096b rsa demo. ) m - in same - is 4096b wide.
asciilifeform: compute 'barrettoid' c for the current m : c := 2^4096 / p . this is done via the knuthian longdividotron, once per modexp.
asciilifeform: note that this is a 4097-bit ( in actual ffa, 8192b ) constant.
asciilifeform: now , for each reduction,
asciilifeform: compute the quotient-guess q, q := (c * x) >> 4096
asciilifeform: if q' is the actual quotient, the inequality q' <= q <= 4q' holds. ( proof left for the reader )
asciilifeform: so now we gotta r := x - q*m , from 1 to 4 times ( this is done via subtract and mux-on-carry , so it always happens 4 times, but when q becomes == to q' the mux stops changing the result )
asciilifeform: now r is == to the remainder of x/m .
mircea_popescu: sounds like the whole discussion coupla weeks ago actually found its way into actual practice.
mircea_popescu: significant gains ?
asciilifeform: in ordinary barretting, x loses its bottom however-many bits, to make the multiplication smaller, depending on the 'bitlength' of x; and then its gets restored with a leftshift . we can't do this, because ffa.
asciilifeform: mircea_popescu: this is quite different from the apeloyee d00d's suggested method
asciilifeform: he proposed to do the normalization as usual, but with 'secret shift' algo. which , any way you cut it, eats more cycles than knuthian division even.
asciilifeform: as for gains, i dun know for sure yet, because i gotta manually implement the asymmetric karatsubaization for the above
mircea_popescu: asciilifeform both salient points (the multi-try mod ; the divide powers of two by p) were part of a larger discussion where you defended tooth and nail teh impossibility!
asciilifeform: given that currently we have only symmetric (equal-bitness multiplicands) multiplication, which will waste massive cycles
asciilifeform: mircea_popescu: i did conjecture that you can skip normalization if you widen the register.
asciilifeform: ( at the cost of n log n )
mircea_popescu: hey, i'm not complaining.
mircea_popescu: on teh contrary
asciilifeform: the (unsurprising) surprise is that ^method is wholly absent from the public lit
asciilifeform: until nao.
asciilifeform: << i oughta expand on this step : first we do r := x - q*m ; then it needs from 0 to 3 further subtractions of m ( we do this muxically . )
a111: Logged on 2017-10-05 16:10 asciilifeform: so now we gotta r := x - q*m , from 1 to 4 times ( this is done via subtract and mux-on-carry , so it always happens 4 times, but when q becomes == to q' the mux stops changing the result )
asciilifeform: anyway that's the whole thing.
deedbot: << Trilema - The problems of today
asciilifeform: i posted it in case meteorite falls on asciilifeform some time in next wk.
trinque: hm no a111 quote?
trinque: mircea_popescu: ^
mircea_popescu: trinque bots are set to ignore bots currently.
trinque: ah k
mircea_popescu: does it reference like that automatically ?
trinque: shall yes, though there are still a few human steps in there today.
trinque: but all the stuff works, which is neat
trinque: I'm in the position to boop a qr code on an airgap machine for the signed tx
mircea_popescu: maybe it's legitimate to except deedbot from "bots" then, seeing how it'll never read loglines which is what the ocncern was.
trinque: yep, wont
mircea_popescu: phf ben_vulpes Framedragger excepting deedbot from the "don't read out bods" ? feasible ?
asciilifeform: !~seen Framedragger
jhvh1: asciilifeform: Framedragger was last seen in #trilema 11 weeks, 3 days, 22 hours, 53 minutes, and 20 seconds ago: <Framedragger> yeah.
asciilifeform: d00d lost at sea or wut.
mircea_popescu: he said something about a new job or moving or something, i'm foggy on details.
asciilifeform: iirc it was a move out of britanistan.
a111: Logged on 2017-07-14 16:39 Framedragger: [PSA, framedraggerish duties to resume soon and answers to irc pings to come (country move done nao, much win). talk soon]
mircea_popescu: maybe he got a new gf or something.
mircea_popescu: by the way, since we're doing PSAs : how about NOT doing dumb shit like this, as a matter of well maintained policy, so we have a chance in hell of knowing whether you've been fucking kidnapped as opposed to you know, merely unreliable in the family way.
asciilifeform: hopefully new gf is not that isidora isis whorecraft or what was her name
mircea_popescu: lol why not ?
asciilifeform: usg mata hari ?
mircea_popescu: if that means she'd at least put some fucking doing into the fucking, i'm all for it.
asciilifeform: the radio silence does suggest -- as mircea_popescu pointed out earlier -- that the fucking went in wrong direction.
mircea_popescu: eh, suggests nothing. what does it suggest in mike's case ?
mircea_popescu: !~seen mike_c
jhvh1: mircea_popescu: mike_c was last seen in #trilema 3 weeks, 3 days, 13 hours, 38 minutes, and 44 seconds ago: <mike_c> congrats! I did that a long time ago too. Good luck with it.
asciilifeform: iirc when mike_c vanished the 1st time, all possible explanations were on the table.
a111: Logged on 2017-08-13 20:36 mike_c: i know, right. It's frustrating prioritizing time. Didn't have time to hang, definitely not to work, so felt weird dropping in once a month and saying "hey, haven't done anything, don't have time to do anything"
mircea_popescu: and then...
trinque: Framedragger appears to be still connected, so maybe he'll check his bouncer and chime in.
asciilifeform: lol maybe mike_c found his new FG's to be a downer
trinque: if folks are ashamed of not having time for the republic, there's more shame in being unable to say so.
asciilifeform: trinque: mike_c's bouncer stayed up for 3+yr
trinque: aha
mircea_popescu: or maybe he just cashed in whatever was left and paid usg for a new mortgage, or maybe was actually shot and left to bleed somewhere, or or or.
mircea_popescu: that's the problem with mystery -- it's mysterious.
asciilifeform: nacht und nebel.
cruciform: is it OK to use waybackmachine for FG instructions?
mircea_popescu: there's prolly an version too
asciilifeform: i refreshed of all the major pages 2-3 day before dulap went
cruciform: thanks
asciilifeform: ( incl . /phuctored , /dupes , etc )
asciilifeform: lemme guess, cruciform , you didn't get the paper manual ?
cruciform: asciilifeform: only the device itself
asciilifeform: mircea_popescu: he bought secondhand iirc
mircea_popescu: is this the first case of an actually molested package ?
cruciform: asciilifeform: correct
cruciform: didn't get hit with import tax, unlike diana_coman
asciilifeform: cruciform: any idea re why not ?
cruciform: asciilifeform: rarely happens in UK - maybe 5% of my packages get plundered
asciilifeform: i suspect that packages from amazon, etc. tend to get left alone.
asciilifeform: plunder focuses on the human ones.
mircea_popescu: so far observed rate is 66%
asciilifeform: cruciform: << 26 sept.
cruciform: asciilifeform: tyvm - reading through atm
asciilifeform: mircea_popescu: could easily be a matter of weight. iirc each of the prev. uk purchasers, got >1
mircea_popescu: i guess so
cruciform: asciilifeform: I had 2 delivered
mircea_popescu: maybe they googled both, one seemed like a corp.
mircea_popescu: (laugh if you will, but about 99% of fiat "government" business consists of google and wikipedia searches)
asciilifeform: i dun think customs monkeys get terminals
mircea_popescu: what's this, like soviets that don'\t get uniforms ?
mircea_popescu: the terminals are the fucking point.
asciilifeform: nah, simply for same reason street sweepers don't -- if you give'em terminals, ~0 packages will get stamped
mircea_popescu: afaik they also do.
asciilifeform: then allbetsareoff and i've nfi
mircea_popescu: where;'s that lulz about bloomberg's "beat cops need catpics, costs 200mn, throwaway in 3 years" nobody-goes-to-jail-for-buying-terminals thing ?
a111: Logged on 2017-09-05 13:25 asciilifeform: meanwhile, in monkeystan, >> 'The NYPD has to scrap the 36,000 smartphones it gave cops over the past two years because they’re already obsolete and can’t be upgraded, The Post has learned. The city bought Microsoft-based Nokia smartphones as part of a $160 million NYPD Mobility Initiative that Mayor Bill de Blasio touted as “a huge step into the 21st century.”'
mircea_popescu: dja understand that had they paid 1k per phone, which a) the ipad never cost and b) volume discounts, they'd have justified less than ONE QUARTER of what they actually spent.
asciilifeform: overhead!11
mircea_popescu: 36 k smartphones at the cost of 4444 dollars each.
BingoBoingo: <asciilifeform> i dun think customs monkeys get terminals << Menial USG employment implies terminal for fucking off
mircea_popescu: only suckers pay retail. ustards pay something else entirely.
asciilifeform: toilet seat in 1980s in pentagon, famously, 5k usd ea.
asciilifeform: mircea_popescu: iirc 'sectera' nsa-issued pnoje, is quoted at 7k ea.
asciilifeform: ( issued to nomenklatura )
mircea_popescu: what, nypd ?
asciilifeform: nah that one's in usg proper
asciilifeform: iirc state officials, even including governor, dun get crypto.
mircea_popescu: crypto ?
mircea_popescu: lolz. spare me.
asciilifeform: btw re nypd, oct. 23 2014 : 'Thanks to a massive $160 million investment, the New York City Police Department is on its way to receive a combination of up to 41,000 smartphones and tablets. Known as the NYPD Mobility Initiative, which will be mostly financed by criminal asset funds provided by the Manhattan DA's Office, the goal is to provide the the city's law enforcement with tools that can improve and streamline their overall work
asciilifeform: flow...'
BingoBoingo: !~ticker --market all
jhvh1: BingoBoingo: Bitstamp BTCUSD last: 4303.97, vol: 7632.28449798 | Bitfinex BTCUSD last: 4310.3, vol: 23219.84356935 | BTCChina BTCUSD last: 4234.9632, vol: 0 | Kraken BTCUSD last: 4307.4, vol: 3031.971074 | Volume-weighted last average: 4308.61469396
BingoBoingo: !~bcstats
jhvh1: BingoBoingo: Current Blocks: 488425 | Current Difficulty: 1.123863285132E12 | Next Difficulty At Block: 489887 | Next Difficulty In: 1462 blocks | Next Difficulty In About: 1 week, 4 days, 22 hours, 37 minutes, and 23 seconds | Next Difficulty Estimate: None | Estimated Percent Change: None
mircea_popescu: asciilifeform obviously, you didn'[t expect the actual USG has any resources ?
asciilifeform: ( sores: and elsewhere )
mircea_popescu: either they use your money or else they don't get shit.
asciilifeform: from where else does crown get resources, lol
asciilifeform: it dun sow or reap
BingoBoingo: It also discourages planting and reaping generally
asciilifeform: revisiting upstack, suggests that i'ma have to go back to the old karatsubatron, which did not insist on powers-of-two bitnesses
a111: Logged on 2017-10-05 16:07 asciilifeform: note that this is a 4097-bit ( in actual ffa, 8192b ) constant.
asciilifeform: ( we're talking about a 3fold speed difference )
asciilifeform: will need asymmetric comba, too...
asciilifeform: continuing , this is not an ordinary mult, we only want the upper half of the bits .
a111: Logged on 2017-10-05 16:08 asciilifeform: compute the quotient-guess q, q := (c * x) >> 4096
mircea_popescu: asciilifeform you could just make it 4096. p can never be 1.
asciilifeform: recall, whole point is to make modexp that worx for all inputs
asciilifeform: otherwise we'd use montgomerization and call it a day
mircea_popescu: but not allowing 1 as a divisor is not some sort of big deal.
asciilifeform: ( i actually did consider, 'special case for x = 1 ' but now you've broken the constanttimealwaysandforeverness )
mircea_popescu: perfectly fine definition of arithmetics, "+0 and /1 are not permitted ops"
asciilifeform: going for ordinary arithmetics in ffa, not oddball alt-arithmetic.
asciilifeform: ( can you prove that cramer-shoup, or whatever we go with next yr, worx in yer alt-arithmetic, lol )
mircea_popescu: in so many words, yes. arithmetic as it is does not allow either op.
asciilifeform: modulo 1 is a permitted op
asciilifeform: yields 0.
asciilifeform: ( modulo 0 is disallowed, it is same as a div0 )
mircea_popescu: << consider that part. in order for 2+3 to be unique, it has to unpermit 2*1*1+0+3*1+0+0+0+0
a111: Logged on 2017-09-20 20:27 asciilifeform: how about the 'every int has unique prime factorization' item ?
asciilifeform: bad enuff that we gotta check every single motherfucking unknown integer for equality to 0, before feeding it into division or modulo
asciilifeform: mircea_popescu: 'holes' in arithmetic suck, because they are inescapably branch conditionals. if i could not prove that div0 is impossible in the course of, e.g., modexp intermediate steps, ffa would be a wholly impossible thing.
mircea_popescu: division by 1 as part of modexp is not happening either.
asciilifeform: it is if defined over integers.
mircea_popescu: your ffa does reals modexp ?
mircea_popescu: i had nfi 2+x modulo 7.5 is even defined.
asciilifeform: as in, i get to , e.g., r = x mod ( a - b ) for some a, b
asciilifeform: reals dun exist on a comp
mircea_popescu: well so then. yes over fucking integers.
asciilifeform: aha, integers. not integers-with-thesehere-holes.
mircea_popescu: 1 is not a divisor of anything, whadda ya want from me.
asciilifeform: 1 is a possible output of a - b. or whatever other combinatorics.
mircea_popescu: so is 0.
asciilifeform: ( as is 0, and this costs us, because you gotta have the div0 trap )
asciilifeform: however div0 trap cannot leak timing, because it is a fatal condition
asciilifeform: terminates the program.
mircea_popescu: im not sure even this much is correct.
mircea_popescu: in that it terminating the program at a certain time it leaks... a certain time.
asciilifeform: a div0 does not have a defined result, and any program that attempts it is resultingly nonsense. the ancients got this right.
asciilifeform: there is no way around this, if you proceed after div0, you are in neverneverland. fortunately writing pcode where div0 demonstrably never happens, is not difficult.
mircea_popescu: and writing code where division by 1 also never happens is an exercise in impossibility ?
asciilifeform: quite nearly.
asciilifeform: i'ma leave this as exercise, and if it stays unobvious, demonstrate laters.
asciilifeform: meanwhile : >> 'Russian Hackers Stole NSA Data' lolz
mircea_popescu: must be all them encrypted derpphones
asciilifeform: '...identifying the files through the contractor’s use of a popular antivirus software made by Russia-based Kaspersky Lab, these people said.' << lol re smear
a111: Logged on 2017-08-18 23:01 asciilifeform: 'That followed the General Services Administration removing Kaspersky from an approved-vendors list in early July and a congressional push to pass a law that would ban Kaspersky from being used by the Department of Defense.'
mircea_popescu: and in other lulz, today we've seen 90 blocks in 18 hours. bitcoin working at half-capacity.
mircea_popescu: !!up apeloyee
deedbot: apeloyee voiced for 30 minutes.
asciilifeform: !!rated apeloyee
deedbot: asciilifeform rated apeloyee 1 at 2017/09/13 18:25:08 << barrett reductionist
asciilifeform: oughta be self-powered
apeloyee: !!up
deedbot: You must be registered.
apeloyee: see?
mircea_popescu: apeloyee is your nick registered with nickserv ?
apeloyee: I did register with nickserv, and identified to it just recently
a111: Logged on 2017-10-05 16:06 asciilifeform: compute 'barrettoid' c for the current m : c := 2^4096 / p . this is done via the knuthian longdividotron, once per modexp.
asciilifeform: modulus
apeloyee: did you mean 2^8192?
asciilifeform: work it out on paper.
asciilifeform: but yes, misprint, the 'p' oughta read 'm'
asciilifeform: range of m is 1 .. (2^4096) - 1 in this example.
asciilifeform: of x is 0 .. (2^8192) - 1 .
asciilifeform: of r, consequently, same as of m .
apeloyee: if q' <= q <= 4q' , it means you don't even know ONE significant digit of q'
asciilifeform: ^ also misprint, good catch
asciilifeform: meant to say that it differs by from 0 to 4 multiple of m .
asciilifeform: from the actual quotient.
asciilifeform: ( just like in ordinary barrett . )
apeloyee: ok, let m =2^4096 - 1. then c=1
asciilifeform: c cannot equal 1 !
asciilifeform: 2 ^ 4096 is a 4097-bit number
ben_vulpes: << pls to update , i forgot this log-spec-addenda; will implement.
mimisbrunnr: Logged on 2017-10-05 16:14 mircea_popescu: phf ben_vulpes Framedragger excepting deedbot from the "don't read out bods" ? feasible ?
mircea_popescu: good idea.
apeloyee: << c := 2^4096 / p; i thought you only have integer division?
a111: Logged on 2017-10-05 16:06 asciilifeform: compute 'barrettoid' c for the current m : c := 2^4096 / p . this is done via the knuthian longdividotron, once per modexp.
asciilifeform: we have knuthian division, it gives quotient and remainder
apeloyee: ok, so quotient of division of 2^4096 by 2^4096-1 is 1.
ben_vulpes: << i have acquired a few afk crises which have eaten a bunch of time since the move.
mimisbrunnr: Logged on 2017-10-05 16:26 trinque: if folks are ashamed of not having time for the republic, there's more shame in being unable to say so.
mircea_popescu: ben_vulpes how about now ?
mircea_popescu: !!help sexpr
asciilifeform: actually apeloyee << strike that, 8192 is correct
a111: Logged on 2017-10-05 18:09 asciilifeform: 4096.
asciilifeform: i'ma stop trying to talk in 2 terminals nao, brb
jhvh1: mircea_popescu: (help [<plugin>] [<command>]) -- This command gives a useful description of what <command> does. <plugin> is only necessary if the command is in more than one plugin. You may also want to use the 'list' command to list all available plugins and commands.
mircea_popescu: !~help sexpr
jhvh1: mircea_popescu: Error: There is no command "sexpr".
mircea_popescu: !$help json
asciilifeform: apeloyee was right in that c is always 2^oneplusbitnessofx / m .
ben_vulpes: bot audit
ben_vulpes: mircea_popescu: ty
lobbesbot: mircea_popescu: (help [<plugin>] [<command>]) -- This command gives a useful description of what <command> does. <plugin> is only necessary if the command is in more than one plugin. You may also want to use the 'list' command to list all available plugins and commands.
apeloyee straps on gas mask
mircea_popescu: ok so results of audit : Framedragger, shinohai, phf : your bot has no help implemented whatsoever, in spite of spec. trinque Framedragger you don't follow the json/sexpr portion, bot simply puts out the same help.
mircea_popescu: now the question here is, do we actually want to ditch the machine portable portion of the help ?
apeloyee: asciilifeform: ok, how it's different then from
a111: Logged on 2017-09-20 21:18 apeloyee: asciilifeform:
mircea_popescu: ben_vulpes because we just don't care about machine discovery, now or ever ?
ben_vulpes: i am struggling to see the case of ever. someone programming against bots should read the help themselves and implement the api described.
mircea_popescu: well the original case was to end up with a proper universal api.
mircea_popescu: exactly like the original url worked, in practice the spec never delivered this magical ability to query servers for resources to the end user.
mircea_popescu: but practice aside, shouldn't i have the ability to query a bank of seated whores, "what can you do ?"
ben_vulpes: should be binary, #t-spec compliant or not.
mircea_popescu: i'm sorry. what should ?
apeloyee: and ftr secret shift does work. if asciilifeform thinks it's slower than division, he must be chugging really strong stuff.
ben_vulpes: the response from the bots to what they can do. either they're all spec-compliant, respond the same way to the same set of commands, or they don't.
mircea_popescu: well yes. the part where the spec should be followed is not under discussion.
ben_vulpes: and what, a machine-understandable description of what each extra-spec command does? i didn't think anyone here operated machines that /thought/.
mircea_popescu: the part what's under discussion is the ~why~ nobody implemented the json/sexpr part of the spec./
ben_vulpes: sloth, likely.
mircea_popescu: ben_vulpes technology is built by steps. so far, being able to put in machine terms a "ask all the bots see which one mentions 'payments' as a string" is a legitimately useful ability.
mircea_popescu: why can't i grep "hurr" over a cat #trilema bots ?
asciilifeform: apeloyee: didja deliberately ignore my observation where multiword shift is intrinsically different timing than subword ?
asciilifeform: your secret shift leaks timing, apeloyee
ben_vulpes: mircea_popescu: okay, i see it.
ben_vulpes: will bring bot to spec.
apeloyee: leaks _what_? it does the same fixed shifts, regardless of operands
asciilifeform: a shift by multiple of the wordsize takes considerably less time than by not
asciilifeform: and to make this untrue, you gotta do W (bitness of ffa) shifts by 1, at all times.
asciilifeform: which is same work as knuthdivide.
apeloyee changes gas mask canister
apeloyee: a shift by multiple of the wordsize takes considerably less time than by not << and a multiply takes more time than add, what.
mircea_popescu: apeloyee but now you can tell what size the shift was, to some degree, by timing.
mircea_popescu: !!up apeloyee
deedbot: apeloyee voiced for 30 minutes.
apeloyee: the operations themselves leak anyway. i thought they shouldn't leak operands?
mircea_popescu: size of shift is an operand
apeloyee: secret shift does FIXED. SHIFTS.
mircea_popescu: trinque << any idea what this is ?
a111: Logged on 2017-10-05 18:04 deedbot: You must be registered.
asciilifeform: the one posted by apeloyee -- leaks
apeloyee: (except the first sub-word one)
mircea_popescu: well excepts are not good signs are they ?
a111: Logged on 2017-09-21 16:22 asciilifeform: !~later tell apeloyee i studied your algo, it (aside from truly massive cost, that would annihilate savings from newton, or barrett, or just about any other trick) ~still leaks~, because shifting by >wordsize is a fundamentally different op from shifting <wordsize; and the only way for this to not be true is for all shifts to happen as a series of wordsize shifts; and a shift by ffawidth-1 (max shiftness) would then consist of ffawor
apeloyee: log(bitness) passes over and the same number of muxes are "massive costs" in your universe?
trinque: apeloyee: you tried sending a private message to deedbot with !!up ?
apeloyee: asciilifeform: write to gmp authors that you got your 4096-bit mul faster than that.
apeloyee: trinque: yes.
asciilifeform: apeloyee: it ain't log(bitness) ! not if you don't want to leak any info re the shift amount.
apeloyee: i'm all out of gas mask cans! HOW does a FIXED shift leak?
asciilifeform: << apeloyee's 'secret shift'
asciilifeform: let's do it in detail
asciilifeform: there are 2 invocations of conventional ffa shift :
asciilifeform: ln 36 : A := Shift(A, S mod W);
apeloyee: yes, the sub-word shift
asciilifeform: ln 41: Scratch2 := Shift(A, S1);
apeloyee: the whole-word shift
apeloyee: *whole-number-of-words
asciilifeform: first of all, didja ever say how to dispose of the 'while' statement ?
apeloyee: fix word-size and bitness. otherwise, you don't
asciilifeform: second, ~any~ wholeword shift leaks info re the shift amount, because different address sequence .
apeloyee: goddamnit. it's FIXED!!! first sub-word one (which doesn't leak in our model), then by 1 word, then by 2, then 4, 8 and so on
asciilifeform: hm i think i finally see what algo apeloyee was trying to implement ( the pseudocode doesn't actually do it )
asciilifeform: you shift by all possible whole word shifts, and mux-keep the correct one; then shift by all wordsize-1 possible subword shifts, and muxkeep the right one.
asciilifeform: this is in fact cheaper than knuth div.
apeloyee: grrr
mircea_popescu: a hey that's actually an idea.
mircea_popescu: apeloyee why not ?
asciilifeform: not only an idea, but afaik the only practical method that isn't W-shifts-by-1-with-mux
mircea_popescu: and it is in fact log-ish W
apeloyee: I do log2(bitness/wordsize) shifts
apeloyee: not number-of-words
asciilifeform: apeloyee: Shift(A, S mod W) leaks S mod W on machines without a barrel shifter.
apeloyee: I assumed the elementary shift doesn't leak. whatever, just do the sub-word shifts with the same algorithm (with word size 1)
apeloyee: for exactly log(bitness) shifts and muxes.
asciilifeform: is what i described above neh
asciilifeform: nonleaking variant of apeloyee's algo.
asciilifeform: and i suspect that he is right that it would beat doublewide-x variant.
apeloyee: wrong. not wordsize-1 shifts; log2(wordsize) shifts.
mircea_popescu: apeloyee no ints are ever exactly log(something) :D
asciilifeform: he means bitness-of-wordsize - number of shifts
mircea_popescu: i'm just being a pain.
apeloyee: per spec the bitness is a power of two
trinque: apeloyee: gpg: BAD04B14A4545828FABCE63C3DB30625393C0BB1: skipped: unusable public key << gpg has this to say about your fp
mircea_popescu: !!key apeloyee
mircea_popescu: trinque well in fairness, it does say BAD right there in the leadup
trinque: lol
trinque: case closed
mod6: :D
apeloyee: I did up myself once, and it worked
asciilifeform: it imports here
asciilifeform: gpg1.4.10
mircea_popescu: trinque you got corrupted, key is fine
asciilifeform: d00d has perfectly working pgpkey.
trinque: yes, and I barfed the key out of same gpg to put it on the wot site
asciilifeform: oh hey it's a clock item
asciilifeform: trinque's clock is ahead of mine
asciilifeform: and key expired.
asciilifeform: ( in trinque's planet, but not yet here )
asciilifeform: why didja make a key that expired in weeks, apeloyee ?!
apeloyee: hm, right. well...
mircea_popescu: expires 0 ?
apeloyee: the key expired.
asciilifeform: subkey, to be specific.
mircea_popescu: apeloyee what was the thinking there ?!
apeloyee: no thinking that i can think of.
asciilifeform: apeloyee: please make a new one ? and give to trinque , signed with old
asciilifeform: and possibly he will agree to swap it in.
trinque: when I get around to it sure
trinque: but this is getting ridiculous
asciilifeform: ftr apeloyee has entirely valid, imho, algo, but terribly confusing pseudocode
asciilifeform: !!up apeloyee
deedbot: apeloyee voiced for 30 minutes.
asciilifeform: though it does look like it'll have to do log2(wordbitness) shifts for the initial subword slide, rather than 1, to avoid leaking on machines without barrel
asciilifeform: ( example : if wordbitness is 64, any subword shift is from 0 .. 63, and therefore can be expressed using 6 muxgated shifts . )
asciilifeform: ( first : 'maybe shift by 1' , second : 'maybe by 2', last : ' ... by 32' )
apeloyee: so do you agree it's O(bitness log bitness) now?
asciilifeform: same thing we do for the wholeword shift
apeloyee: painful, but probably beats a larger muliply
asciilifeform: it gotta
asciilifeform: ty for properly rubbing it in asciilifeform's nose, apeloyee
apeloyee: btw, i'd like to know the answer to
a111: Logged on 2017-10-05 18:20 apeloyee: asciilifeform: ok, how it's different then from
asciilifeform: apparently ~nobody else properly reads.
asciilifeform: apeloyee: it won't differ at all, once i have working shifter using your scheme.
asciilifeform: as i understand will be quite exactly it.
a111: Logged on 2017-10-05 16:15 asciilifeform: the (unsurprising) surprise is that ^method is wholly absent from the public lit
asciilifeform: i was discussing a 'shiftless' doublewide barrett
apeloyee: so was I
apeloyee: in that link
asciilifeform: apeloyee: what is B ?
apeloyee: 4^K = R*N + B, 0 <= B <= N (line seven). pseudo-remainder.
apeloyee: if N is 2^(K-1), then ordinary quotient won't fit in K+1 bits. but pseudo-quotient (one less the actual quotient) still works.
apeloyee: meaning you divide 4^K - 1 by the modulus, not 4^K
asciilifeform: why 4^k ?
apeloyee: we had that thread today!!!
asciilifeform: aa 2^(2k) ok
asciilifeform: apeloyee: so what'd the correction step look like, with the pseudoquotient
apeloyee: one subtraction and mux.
asciilifeform: oh hm this worx.
asciilifeform: adds exactly one sub-and-mux to the existing 3.
asciilifeform: neato apeloyee , ty.
apeloyee: why three?
asciilifeform: was 3 in barrett's original proof. but with doublewide c, looks like only 1 ? ( though i do not yet have a proof for this )
apeloyee: in the posted version exactly ONE is needed
apeloyee: barrett original needed two
asciilifeform: i can read, and see this. but why is it 1 ?
apeloyee: A - N*floor(A*R/4^K) < 2*N <- do you agree with this? (the proof is unnecessarily complicated in that paste)
asciilifeform: dunno if i do, still trying to swallow the proof.
apeloyee: what step is a problem?
asciilifeform: let's start with the above
apeloyee: let rewind. A*R = A*(4^K - B)/N = (A*4^K)/N - A*B/N; as B <= N, then the previous >= (A*4^K)/N - A. clear?
apeloyee: A <4^K, so the above strictly >(A*4^K)/N - A.
apeloyee: which is equal to (A/N - 1)*4^K
apeloyee: ie our result is off by at most one
asciilifeform: ok indeed worx.
apeloyee: and in one side only, as obv can't be bigger than (A/N)*4^K
asciilifeform: !!up apeloyee
deedbot: apeloyee voiced for 30 minutes.
asciilifeform: !!rate apeloyee 2 A - N*floor(A*R/4^K) < 2*N
asciilifeform: !!v 17BC89B25DD39E071B293EDE353D50C10A165F320E436ED7FE36D996140DBBCC
deedbot: asciilifeform updated rating of apeloyee from 1 to 2 << A - N*floor(A*R/4^K) < 2*N
asciilifeform: eagle eyes, apeloyee . i dun suppose you have a constant time gcd up your sleeve ?
apeloyee: now let's try generalizing to standard barrett (the error will grow to 2 of course). let L be number of digits in N: 2^(L-1) <= N < L. L is calc'd with the CLZ algorithm
apeloyee: what you need a gcd *for*
asciilifeform: miller-rabin
apeloyee: I don't see any GCD in miller-rabin
asciilifeform: for the initial sieve ~prior~ to miller-rabin
asciilifeform: want to gcd(candidate, biggestprimorialthatfitsintheffabitness)
asciilifeform: save massive time
asciilifeform: the other obvious place for gcd is rsa phi
apeloyee: asciilifeform: just do trial division.
asciilifeform: and in who knows what future numbertheoretical ops. we gotta have gcd.
asciilifeform: trial division is dog slow vs gcd.
apeloyee: you gcd is O(N^3), and so is trial divisiom
apeloyee: *your
asciilifeform: non-constanttime gcd is certainly not O(N^3)
asciilifeform: why do you think constanttime gotta be, apeloyee ?
asciilifeform: take lehmer's gcd as a starting point.
apeloyee: I'm simply saying trial division is better than what you have
asciilifeform: i dun have a working gcd yet
apeloyee: you proposed a euclidean one.
asciilifeform: euclidean'd be o(n^3) yes
asciilifeform: ( and still not constant time )
apeloyee: just do some muxes in the end. 2*bitness divisions obv suffice (actually less, but I'm sleepy now)
asciilifeform: currently i suspect that it is possible to constantize lehmer's logn gcd.
asciilifeform: ( or not ? )
asciilifeform: hm, i must've been thinking of some other gcd algo, because lehmer's is a definite dead end.
apeloyee: but here's an O(n^2 log n), for a large value of constant.
asciilifeform: apeloyee: if you can think of a subquadratic constanttime gcd, plz do write in
apeloyee: well, not subquadratic :( but obv correct
asciilifeform: let's have it
apeloyee: let a, b be inputs; a > b. shift b left so that it has one digit less than a (using CLZ and secretshift), subtract b from a repeatedly (at most thrice). b:= new result, a:=b
asciilifeform: what happens if a <= b ??
apeloyee: you sort them at the biginning, using a mux.
asciilifeform: so i sorted them, answer is for the sorted pair then, not the actual pair neh
asciilifeform: and say they're equal.
asciilifeform: what then ?
asciilifeform: other thing is that i dun see how this is constant time
asciilifeform: time will vary massively depending on hamming weights
apeloyee: on what planet gcd(a,b)<>gcd(b,a).
asciilifeform: q remains tho, how to handle a==b case
apeloyee: if they're equal -that's the end, the gcd is found. don't throw out the unshifted b.
asciilifeform: that ain't constant time !!
asciilifeform: thing has to take exactly.same.number.of.steps. regardless of what a,b equal.
apeloyee: do what you'd do in euclidean.
asciilifeform: pretty sure you just described lehmer's gcd, actually
apeloyee: grind until you did (2*bitness) steps. since remainder can't be more than a/2.
BingoBoingo: In other prohetics found in the prophecy mines: "In the entire Southeast Asia, the natives know quite well that the Western expats are nothing but biatches with no integrity whatsoever. Back in their own home countries, the Asians are dominating them in every professional fields, creating various gigantic ethnic enclaves and making them Westerners becoming mere simple cheap tourists. Whenever the Westerners visit any of the Asian-created
BingoBoingo: ethnic enclaves, they never fail to put on a smile and admire everything Asian. Without the Asian industriousness and food, the Westerners will just be cave people just like their cave-dwelling ancestors not too long ago. One has to wonder why didn’t the Mongol overlords finish colonizing these Westerners and turned them into proper Asians. In the near future, the Chinese will just do that because the spineless lazy Westerners have gr
BingoBoingo: own to be dependent on everything Chinese now."
asciilifeform: apeloyee: maybe i'm thick, but i don't see how to grind 'shift so b has one digit less than a' a fixed number of times for all possible a and b.
asciilifeform: can i persuade you to write this out as algo , apeloyee ?
apeloyee: you shift (b) left by max(0,CLZ(a)-CLZ(b)-1) bits
asciilifeform: yes that's one shot
asciilifeform: but what are the fixed shots ?
asciilifeform: and on what does their number depend.
apeloyee: 2*bitness shots suffice (actually less). see above .i'll write it but not now.
a111: Logged on 2017-10-05 19:54 apeloyee: grind until you did (2*bitness) steps. since remainder can't be more than a/2.
asciilifeform: ok laters then
apeloyee: bye.
trinque: !~later tell apeloyee glad to see someone grinding the ffa forward with asciilifeform. get me that fixed key!
jhvh1: trinque: The operation succeeded.
asciilifeform: !~later tell mircea_popescu << possibly lulzy: garza convicted
jhvh1: asciilifeform: The operation succeeded.
asciilifeform: tldr: usg inquisitor imposes 10mil $ fine, plus whatever jail ( decided later )
asciilifeform: !#s garza
BingoBoingo: Scum scuming scum
a111: Logged on 2017-10-05 19:20 apeloyee: we had that thread today!!!
mircea_popescu: asciilifeform pretty sure that's BingoBoingo o no wait, was it cazalla's good friend ?
mircea_popescu: the guy with the slutty wife.
asciilifeform: waiwat who
mircea_popescu: homero garza
mircea_popescu: ran a fake miner farm scam fore a while in iirc 2013
a111: Logged on 2017-10-05 19:50 apeloyee: on what planet gcd(a,b)<>gcd(b,a).
mircea_popescu: alf is obv lefthanded, has problems with commutativity.
a111: Logged on 2017-09-20 19:26 asciilifeform: mircea_popescu: lookx to me like your addition dun commute ( or associate... )
mircea_popescu enjoyed reading convo, evidently enough.
asciilifeform currently making an efficient seekritshift, it is somewhat tricky
asciilifeform: incidentally ( and luckily for asciilifeform ) asciilifeform is not left-handed : left-handers were savagely beaten on asciilifeform's home world
mircea_popescu: at age 4 ?
mircea_popescu: << ahahaha what is this bs, some other nobility elliot ?
a111: Logged on 2017-10-05 19:55 BingoBoingo: ethnic enclaves, they never fail to put on a smile and admire everything Asian. Without the Asian industriousness and food, the Westerners will just be cave people just like their cave-dwelling ancestors not too long ago. One has to wonder why didn’t the Mongol overlords finish colonizing these Westerners and turned them into proper Asians. In the near future, the Chinese will just do that because the spineless lazy Westerners have gr
asciilifeform: mircea_popescu: i take it this custom was not alive in classical ro ?
mircea_popescu: asciilifeform not really. they even tolerated left hand writing, to some degree.
asciilifeform genuinely surprised
mircea_popescu: bear in mind tho that i come from the civilised third of romania.
asciilifeform: vlad's dukedom iirc ?
mircea_popescu: no, vlad was voievode of the south, actually. this'd be the northwest.
mircea_popescu: at the time, right and proper austrian lands.
mircea_popescu: whereas most stories of romanian independenciality tend to come from 3rd 3rth, the north east.
mircea_popescu: stephen and such.
asciilifeform: interesting combo, boys fighting over meat in soup, but left-handing -- tolerated
mircea_popescu: which soup is this ?
asciilifeform: hm could've sworn it was in l0gz
asciilifeform: alluded to in but iirc was in a trilema piece
a111: Logged on 2017-02-19 16:12 mircea_popescu: (the source was 80s dispute wrt to general public's apparent bottomless appetite for purchasing luxury goods -- in eastern communism this includes meat, washing machines, electricity, etc -- for which they won't even have the fucking decency to pay in hard currency, but aim to use the worthless reminbo internal scrip)
asciilifeform: where girls in mircea_popescu's school would beg him for meat or how did it go
asciilifeform: did i dream this?
mircea_popescu: boys fought for the proper cut of bread (ie, the end bit, for some reason, was the only proper bit). but the last time communal eating was common was just about kindergarten
mircea_popescu: asciilifeform oh, yeah, but teh story was subtly different.
mircea_popescu: such as -- 4yos ; girls not boys ; mostly me being an uppity snot. that sorta thing.
asciilifeform: this piece was memorable, because at no point was meat any kind of luxury for asciilifeform & co.
mircea_popescu: pretty much all common man could have for the asking wo9uld be dog food parts, lower chicken legs, necks, that sorta thing.
mircea_popescu: this, throughout 80s.
mircea_popescu: romania had a plan!
asciilifeform: d00d evidently did his homework.
BingoBoingo: <mircea_popescu> asciilifeform pretty sure that's BingoBoingo o no wait, was it cazalla's good friend ? << He was the cheif jizz mopper cazalla and self cut teeth on
BingoBoingo: <mircea_popescu> ran a fake miner farm scam fore a while in iirc 2013 << Perhaps something that early but mostly 2014 thing. Complete collapse happened between Xmas and New Years
BingoBoingo: << random response to whiny barangs
a111: Logged on 2017-10-06 00:38 mircea_popescu: << ahahaha what is this bs, some other nobility elliot ?
BingoBoingo: In other 12 hour sagas "The remains of a possible fetus were discovered by a custodian Wednesday in the bathroom of Wirth-Parks Middle School in Cahokia. The coroner arrived at the scene shortly after 8:30 p.m. and carried a small, covered box into the school." << "‘Fetal remains’ found in school bathroom actually a used feminine hygiene pad"
asciilifeform: BingoBoingo: what's a barang ?
BingoBoingo: Foreigner, in a language which apparently features a high degree of mutual intelligibility in its region
asciilifeform: aaah gaijin
← 2017-10-04 | 2017-10-06 →