log☇︎
157100+ entries in 0.092s
asciilifeform: ( so that'd be ~week-long keygen. )
asciilifeform: which means that even a minute-long modexp is theoretically fieldable ( you get ~day-long keygen, and minute-per-4096bits decrypt/encrypt, but this is livable, ancestors lived with much slower hand-cranked otp )
a111: Logged on 2017-07-10 19:33 asciilifeform: sooo a 4096b rsa key takes about a dozen modexp's, on avg, on gpg 1.4.10
asciilifeform: incidentally also worth remembering is that http://btcbase.org/log/2017-07-10#1681224 . ☝︎
asciilifeform: and at any rate the 'f' in ffa -- stays.
asciilifeform: and would not help the private exponent op ( where one would want to leak nothing at all )
asciilifeform: naturally this is for the public exponent op only
asciilifeform: incidentally, if you're willing to leak the height of the last set bit in the exponent, you cut the cost even of naive method above, linearly ( e.g. the typical gpg exponent is 17 bits ) -- but i'd very much rather not
asciilifeform: there are no tables in ffa, and an ffa with a table is pointless ( if you access leaklessly, by chugging whole table each time ( we had thread ) you wipe out the time saving. )
asciilifeform: can't do tables. period. they result in addressing-by-secretbits and leak timing from cache.
mircea_popescu: tbh i liked my "all items are 4096 bit" world way better and wtf is this "multiplication doubles width" bs anyway!
mircea_popescu: precalculating 4k mods of 4 to 8kb single-bit items is a pain in teh arse even in simple space terms, what, ima have 16MB worth of ram/disk dedicated to just this PER mod ?
asciilifeform: anyway this was not a sexy/exciting report, it is just to share what was the baseline ( saddest known algo )
mircea_popescu: but no, this dead branch has been fully explored, nothing here.
mircea_popescu: ME TOO!\
asciilifeform: hey i'd be the first to applaud if mircea_popescu came up with magical algo that reduces cpu work to ~0 !
asciilifeform: and is of the bottle.
mircea_popescu: yes, but as long as you do rsa, a ^ b mod m = either a ^ b or else, a ^ b - m. that's the complete story, there are no other cases. there's specifically no k * m parameter to be explored there.
asciilifeform: i think we had this in detail
mircea_popescu: fine. and the reason your mod does not reduce to "At most one substraction op" is "because i believe the product of two prime numbers 2kb each can in fact be less than 2^4095". that it ?
asciilifeform: gotta nail down this difference, because it makes a difference ( in that you cannot exponentiate-and-then-modulus, universe ain't big enough , hence the invention of mod-exp )
asciilifeform: correct. the latter when using the method depicted in the snapshot.
mircea_popescu: that's neither here nor there.
mircea_popescu: so then! calculating a ^ b mod 2^4096 is 0.26s ; calculating a ^ b mod m is 51.3s. that it ?
asciilifeform: FZ_Exp does not calculate a^b, it calculates a^b modulo ffawidth. this is important difference.
mircea_popescu: cuz your production does not fit the requisite form. produce item in requisite form, will gladly use that.
mircea_popescu: i dunno what you think the behaviour preserves, but i can assure you it doesn't.
mircea_popescu: you will refuse to be useful how many times in a row ?
asciilifeform: and in case it isn't clear, the 51+ is using the slowest known method.
asciilifeform: a ^ b mod 2**4096 , rather than a^b ( which our galaxy hasn't enough atoms to represent for 4096bit a and b )
asciilifeform: ( kochian modexp stops at the last 1bit )
asciilifeform: koch is neither here nor there, was not part of the measurements, nor is possible to compare because it does not actually do the same job
asciilifeform: same WITH divisions, i.e. modular, takes 51.3.
asciilifeform: exponentiation WITHOUT DIVISIONS (i.e. nonomodular, or modulo the set ffawidth, in this case 4096) takes 0.26
mircea_popescu: is it or is it not true a modular exponentiation in current gpg takes, on your chosen machine, 0.26 seconds.
mircea_popescu: asciilifeform so to try and extract actual reality from this : a ^ b mod m takes 0.26 seconds in koch writing ; same a^b mod m takes 51.3 on your box, notwithstanding a ^ b without any modding takes 1s ?
asciilifeform: the corresponding item in the src seen in http://btcbase.org/log/2017-09-17#1715329 , is FZ_Mod_Exp . ☝︎
asciilifeform: ( a*b mod m refers to mod-exp, i thought this was clear )
asciilifeform: read the src plox
a111: Logged on 2017-09-17 12:25 asciilifeform: mod exp for 4096bit operands aint 1 mod, but 8192 of'em, and they're on 8192bit ints
mircea_popescu: http://btcbase.org/log/2017-09-17#1715311 << i'm missing something here. did you compare a MULTIPLICATION (as your text actually said, a * b) with an exponentiation ? ☝︎
jhvh1: asciilifeform: The operation succeeded.
asciilifeform: !~later tell mircea_popescu http://wotpaste.cascadianhacker.com/pastes/HFs6x/?raw=true
asciilifeform: ( and then yes rotted and vanished like every previous fat an' comfortable set ) ☟︎
a111: Logged on 2017-09-17 11:38 mircea_popescu: in this sense, the perpetual "we don't deal with problems, we try to inflate them away" is a forced mistake of the socialist mind, brought about by its deffective tech stack. what the fuck else are they going to do, the only people to which "french culture" in its jimbo diderot reformulation appeals to are first generation in shoes functional illiterate tards.
asciilifeform: http://btcbase.org/log/2017-09-17#1715307 << didja sleep though the century when '1st gen in shoes' folx handed 'cultured' europe its arse on a platter, and then flew into space etc ☝︎☟︎☟︎
a111: Logged on 2017-02-03 05:29 phf: my mom had one of those "papa escaped the purges by burning all paperwork but life got tough when they took our governess" piano teachers, who would lament about being forced to teach peasants and brown folk, not on racial grounds as much as "you have to have a bidet in your house for 3 generations, before you can touch a piano"
a111: Logged on 2017-09-17 11:34 mircea_popescu: the whole fucking POINT of french revolution was "we can make things -- soldiers, arms for those soldiers, culture for their fathers and wives, CHEAPER and MORE ABUNDANT if we chinese it." "but will it be as good ?" "i can't hear you over the kanzure of my superficiality"
asciilifeform: ^ and yes there is an 'obvious' 2x speedup possible and not yet done, even in this naive method
BingoBoingo: ty trinque for the reply fodder http://qntra.net/2017/09/media-and-clergy-orchestrated-civil-disorder-erupts-in-downtown-st-louis-after-acquittal/#comment-109785
a111: Logged on 2017-09-15 23:51 BingoBoingo: So this "Sally Hemmings And The Widower" Fanfic/novel, I'm not finding anyway to work crabapple metaphor into unifying narrative
BingoBoingo: <mircea_popescu> http://btcbase.org/log/2017-09-15#1715002 << pick riper fruit till this gestation completes, what's teh rush. not like anyone seriously cares about either. << Or at least sweeter fruit ☝︎
asciilifeform: incidentally do i have to explain why a modexp op cannot be 'exp, and then take result and mod' ? or is it screamingly obvious .
a111: Logged on 2017-09-17 11:08 a111: Logged on 2017-09-12 21:32 mircea_popescu: and this is potentially recursive, in that if you have a 500 bit number with 300 ones in it, you do the mod for 500 terms which are all a power of 2, throw 200 away, keep the other 300 and add them.
asciilifeform: http://btcbase.org/log/2017-09-17#1715293 << ftr this is a description of egyptian div, a SINGLE mod operation. ☝︎
asciilifeform: it's the same 4096 karatsubamuls, 4096 karatsubasquares, 4096 shifts, 4096 muxes, and 8192 ( once for each mul and sqr output, and ergo 8192-bit wide, and they dominate cost ) divs.
a111: Logged on 2017-09-17 11:13 mircea_popescu: what this means in practice being that "the 4096 bit modulus of a 4096 bit number is obtained through at most one substraction op".
asciilifeform: http://btcbase.org/log/2017-09-17#1715295 << this is screamingly wrong, a 4096bit modulus can be anything, incl. 1 ☝︎
asciilifeform: 1 * 1 mod 1 takes same time etc
a111: Logged on 2017-09-17 11:10 mircea_popescu: (pro tip : since the modulus is 4096 bits, all powers of two up to 2^4095 are unchanged by modding.)
asciilifeform: http://btcbase.org/log/2017-09-17#1715294 << computing an 'unchanged' TAKES SAME TIME as any other, omfg, it's constant time ☝︎
asciilifeform: mod exp for 4096bit operands aint 1 mod, but 8192 of'em, and they're on 8192bit ints ☟︎
a111: Logged on 2017-09-17 11:07 mircea_popescu: http://btcbase.org/log/2017-09-16#1715247 http://btcbase.org/log/2017-09-16#1715243 are these to mean therefore that a mod computed on a 4096 bit int takes 50 seconds ?
mircea_popescu: "it's ok, science will progress so fast nobody will ever wear shoes again!!!". yeah, i'm sure. and then it'll "progress" even faster than that huh.
mircea_popescu: in this sense, the perpetual "we don't deal with problems, we try to inflate them away" is a forced mistake of the socialist mind, brought about by its deffective tech stack. what the fuck else are they going to do, the only people to which "french culture" in its jimbo diderot reformulation appeals to are first generation in shoes functional illiterate tards. ☟︎
mircea_popescu: which is why it's always poor immigrants, and "3rd world" shitholes. cuz if you approximate a non-closed form solution by the number five, you get a certain set of perked ears, and they're going to always be in the back of the class.
mircea_popescu: obviously, it won't be as good, in the obvious sense that you don't need frenchmen to do it, can have it done in yugoslavia cheaper. and by yugoslavia i mean poland which is to say china.
mircea_popescu: the whole fucking POINT of french revolution was "we can make things -- soldiers, arms for those soldiers, culture for their fathers and wives, CHEAPER and MORE ABUNDANT if we chinese it." "but will it be as good ?" "i can't hear you over the kanzure of my superficiality" ☟︎
a111: Logged on 2017-09-12 23:10 mircea_popescu: 1. if you actually want metal kbd, your choice of steel is probably ill advised. i'd try silver instead. heuristicallyt there's a reason gunsmiths and silversmiths were ~the same people i nthe early modern period ; moreover silver has better properties in the range sough.
a111: Logged on 2017-09-13 19:36 asciilifeform: lathe driven moar by musketry than by fine art tho
mircea_popescu: http://btcbase.org/log/2017-09-13#1713770 <<>> http://btcbase.org/log/2017-09-12#1713183 ie these are not as distinct as you'd think. ☝︎☝︎
mircea_popescu: also arguably some of the ugliest tits ever to meet lens.
a111: Logged on 2017-09-13 01:18 mircea_popescu: ok, quick summary : tech pre-injectors was carburators. engines "optimized for racing" got slowly more complex over time. then injection became feasible, and they simplified. now they've recomplexified again.
mircea_popescu: http://btcbase.org/log/2017-09-13#1713368 << for some decent lulz on this topic, nonsensical "rebel w/o cause" 1982 john clark item, "running on empty". featuring oddball ventilation towers on the hood and vague, contradictory hints of female sex slavery. ☝︎
a111: Logged on 2017-09-12 23:52 asciilifeform: http://btcbase.org/log/2017-09-12#1713184 << in ffaworld, a < or > or == comparison is not only a subtraction (O(N)) but another O(N) test for nullity (xor all the words together)
mircea_popescu: http://btcbase.org/log/2017-09-12#1713197 << this is perfectly true, but not actually the problem here. ie, it'd be just as braindamaged even without that. ☝︎
mircea_popescu: what this means in practice being that "the 4096 bit modulus of a 4096 bit number is obtained through at most one substraction op". ☟︎
mircea_popescu: (pro tip : since the modulus is 4096 bits, all powers of two up to 2^4095 are unchanged by modding.) ☟︎
a111: Logged on 2017-09-12 21:32 mircea_popescu: and this is potentially recursive, in that if you have a 500 bit number with 300 ones in it, you do the mod for 500 terms which are all a power of 2, throw 200 away, keep the other 300 and add them. ☟︎
mircea_popescu: because that's fucking scandalous to begin with ; and in no case the discussed http://btcbase.org/log/2017-09-12#1712957 can possibly take that fucking long. ☝︎
a111: Logged on 2017-09-16 15:31 asciilifeform: in other olds ( i dun think i posted this measurement ) the NAIVE modular exponentiator takes 51.3 seconds per 4096b a*b mod m , on the 'standard' test box
a111: Logged on 2017-09-16 15:35 asciilifeform: ( for comparison, a NONmodular exponentiation of same width on same box with same ffa takes 0.26s. )
mircea_popescu: http://btcbase.org/log/2017-09-16#1715247 http://btcbase.org/log/2017-09-16#1715243 are these to mean therefore that a mod computed on a 4096 bit int takes 50 seconds ? ☝︎☝︎☟︎
a111: Logged on 2017-09-15 23:51 BingoBoingo: So this "Sally Hemmings And The Widower" Fanfic/novel, I'm not finding anyway to work crabapple metaphor into unifying narrative
mircea_popescu: http://btcbase.org/log/2017-09-15#1715002 << pick riper fruit till this gestation completes, what's teh rush. not like anyone seriously cares about either. ☝︎
deedbot: http://trilema.com/2017/family-business/ << Trilema - Family Business
mircea_popescu: asciilifeform not directly evident giving it to the state's people's something or the other is such a good idea in any case.
mircea_popescu: would make a splendid show to butcher the lot.
mircea_popescu: fortified oil complex worth like 3rd in the world, and outside of that 20mn or so usg nigglets clamoring about gini indexes.
asciilifeform: ( after which xilinx & altera market will look like cisco's -- tame idjits only )
asciilifeform: mats: lattice is the last remaining non-usg-owned fpga house. lizard gosplan terrified that it might get sold and specs for the high density fpga, opened
deedbot: http://trilema.com/2017/costa-rica-de-noche-y-de-dia/ << Trilema - Costa Rica, de noche y de dia
BingoBoingo: es that the gauntlet in question, which features a "kek" symbol that resembles the "Kekistan flag" popularized by 4chan, was originally created by the game's developers back in June of 2015. Dague says the gauntlet was one of many items in the game that "reference real world art, iconography, typeface, and other design elements" and that "some of the reference imagery featured the simple mirrored chevron shapes found in the finished piec
BingoBoingo: In people who wish they were the ministry of games: "Earlier this week, when it became clear that a gauntlet in Destiny 2 resembled a "Kekistan" flag design that has been repurposed by neo-Nazis, developer Bungie was quick to apologize and work to remove the item from the game. Now, the developer is using a public blog post to try to explain how the symbol ended up in the game in the first place. Community Manager David "DeeJ" Dague writ
BingoBoingo: <mircea_popescu> apparently the whole "cashless society" is working out great for the fucktarded "elite" << Aha, Butthole is what's left to take
BingoBoingo will also have to dig it out
mats: kindly mirror pdf? 'resist.com' ver referenced in log, has been taken down
BingoBoingo: mats: I may be able to put ine together this weekend. Prolly quickest to run pdf through text converter yourself.
asciilifeform: a truly eagle-eyed nitpicker will of course notice that... (r13) gurer pnaabg or fhpu n guvat nf abazbqhyne rkcbaragvngvba va ssn, vg vf nyjnlf zbqhyb ng zbfg gur ovgarff!
asciilifeform: when bitness is B, a modular exponentiation takes B mod-muls and B mod-squares, each of which produces a 2B-wide item that gets div'd . that's 2B 2B-wide divisions.