log☇︎
156500+ entries in 0.097s
a111: Logged on 2017-09-20 22:09 asciilifeform: this worx. but seems to reduce to a more complicated and non-unrolled version of http://btcbase.org/log/2017-09-20#1716343
asciilifeform: ( which likewise shifts for the entire bitness, every time )
asciilifeform: i can say immediately that if you 'secret shift' every time you want to modulus, you will take at least same number of steps as knuthian division.
apeloyee: but they are separate things
apeloyee: asciilifeform: the paste contains TWO procedures: CLZ and secret shift. together they solve the normalization problem
a111: Logged on 2017-09-20 21:54 asciilifeform: e.g., N is a number to shift, Q is shift amt, R is result, S = l_shift(N, 1); mux(S, R, R, iszero(Q - 1)); S = l_shift(N, 1); mux(S, R, R, iszero(Q - 2)); ...... mux(S, R, R, iszero(Q - (width-1)))
asciilifeform: this worx. but seems to reduce to a more complicated and non-unrolled version of http://btcbase.org/log/2017-09-20#1716343 ☝︎☟︎
asciilifeform: what part of 'can't branch on bits' was hard to understand ?
asciilifeform: apeloyee: i'd be curious to hear how intel will 'optimize' xor. ☟︎
apeloyee: apologies for the sad parody of pseudocode.
apeloyee: (assuming mux doesn't leak - who's to prevent intel from "optimizing" it?)
apeloyee: asciilifeform: and if you think a little more, you'll notice that log2(bitness) shifts (by wordwidth, wordwidth*2, wordwidth*4 ... ) and muxes suffice. http://p.bvulpes.com/pastes/hkkYj/?raw=true
asciilifeform: fortunately we don't actually need any such thing in ffa. none of the algos require shifting by a seekrit bitness.
asciilifeform: e.g., N is a number to shift, Q is shift amt, R is result, S = l_shift(N, 1); mux(S, R, R, iszero(Q - 1)); S = l_shift(N, 1); mux(S, R, R, iszero(Q - 2)); ...... mux(S, R, R, iszero(Q - (width-1))) ☟︎☟︎
asciilifeform: re shifter, on second thought you COULD shift bitness-1 (e.g. 4095 ) times and then mux EVERY TIME YOU SHIFT but omfg.
asciilifeform: ( normalization LOOKS AT SECRET BITS and therefore is forbidden anywhere in ffa for any reason )
asciilifeform: apeloyee: also if you know how to newton-raphson converge without ANY normalization ANYWHERE, plz to post.
asciilifeform: afaik this cannot be done without a physical barrel shifter of the given bitness. ☟︎
asciilifeform: incidentally if apeloyee or anybody else knows how to make a 1..ffabitness shifter that doesn't leak the shift amount, on ordinary pc, plz post.
asciilifeform: if that's barrett, it'll be barrett. if it's knuthian, will be knuthian. if newtonian -- newtonian. in no case it will be montgomery, say, or any other non-universal one.
asciilifeform: apeloyee: i want the fastest possible universal reduction that works in fixed spacetime and provably so
asciilifeform: ( and when we do rsa, can store the reciprocal in the key, there's no particular reason to compute it every time )
apeloyee: true, but I thought you wanted a universally-fast reduction?
asciilifeform: still 1 knuthian div, vs the 8192 i have now (in the example)
apeloyee: you gotta compute the reciprocal for barrett's somehow.
asciilifeform: because as far as i can see, it has all of the same problems, but more so
asciilifeform: apeloyee: yes because for guarantee of convergence you need MASSIVE ( afaik, at least 4x the ffawidth ) bitness
asciilifeform: they depend strictly on which op ( which is not secret ) and the ffawidth ( also not secret )
apeloyee: asciilifeform: is it intended that ffa doesn't have a shift that doesn't leak the upper bits of shift amount (via access pattern)?
trinque: so she can tutor me with it
trinque: got one that speaks it, actually
mircea_popescu: really great flick to watch with girl ; motivate her to learn french.
trinque: heh, I will be sure to push that one onto the appropriate stack
mircea_popescu: teh republic built the man's dream!
mircea_popescu: tl;dr : guy is practically a gremlin, dedicated and ~meaningfully so~ industrialist, married to a credible "woman's job" a la francaise. they end up quarreling domestically when he destroys her greenhouse to extend his factory ; she runs for mayor against him (he used to own unopposed). his platform is "en premier place, le plein emploi ; en seconde, le plein emploi ; e finalment : LE PLEN EMPLOI!"
asciilifeform: apeloyee: my principal problem with barrett is that i don't have a proof that it works with 3 or fewer corrective-subtractions for ALL possible a,b,m in a*b mod m
trinque: then got the houston flood piece and hypertext-dht piece to write
trinque: anyhow got payments bot to ship, which while running long is forcing me to develop sane processes around airgapped trb use
apeloyee: asciilifeform: how about a "pediwikian bowdlerization of barrett" after all (don't truncate multiply results)? This would be simpler (but slower)
trinque doesn't oppose a guy hand-carving his own thing at all, either.
mircea_popescu: lobbes yes, that'd be an improvement conceptually/
lobbes: re this monitoring thing: seems like the key either way is to have a parent process that spawns the child process, rather than a disconnected process monitoring whatever
trinque: been cranking on being able to cut transactions on an airgap machine with only trb+deps
trinque: it's on teh stack
asciilifeform: from thread some month ago
asciilifeform: hey trinque didja ever have time and write that piece re 'sane http replacement' ?
trinque: and sure, could bash script a loop that does exactly what you want, and then bash script your next service too. people do that. and then they generalize out the utils, and have a process manager of their own. ☟︎
trinque: http://smarden.org/runit/runsv.8.html << stuff the supervisor does
trinque: which can be other than "thrash-loop until operator notices"
mircea_popescu: does it do moar monitoring than "is PID so and so still there" ?
trinque: sure, it knows the PID of its own child process
trinque: and it's the lightest weight item that does so, that I've found
trinque: no, it has a supervisor process that spawns your thing, monitors state, respawns if dead
trinque: grepping for logs to restart is nuts
trinque: eh lobbes that sounds really damned hairy.
lobbes: for the meantime, auctions will work over there an' such
lobbes: I'm gonna look into shinohai's suggestion of checking the PID rather than grepping for name. Let's just leave it in #eulora for now and I'll get back once I dig deeper later tonight/this week
mircea_popescu: lobbes didja give the code a once over for the purpose of not ending up with this again ?
lobbes: I assume IP had to be banned. I'll leave it up to the court whether or not to unban. bot's only function here was quoting logs.minigame.bz anyhow
mircea_popescu: "primes gotta have two digits"
mircea_popescu: heck, for the same money also exclude 2, and perhaps 3 because son of numbers previously hit by bill of numeric attainder.
mircea_popescu: so far i'd be surprised if anyone seriouslty didn't want to ; the argument as presented seems sufficiently overabundant!
mircea_popescu: alrighty then!!1
asciilifeform: just don't be surprised when nobody else wants to.
asciilifeform: go ahead an' bolt whatever you like to ~your~ kalash, hey.
asciilifeform: wtf is even the point of talking about primes if you ALWAYS gotta say 'but not 1'
mircea_popescu: because you gotta add the spit-fix at the right juncture.
mircea_popescu: it does. "unit terms are ignored"
asciilifeform: how about the 'every int has unique prime factorization' item ? ☟︎
mircea_popescu: the definition of prime is as quoted, what more is needed here ? 1 is divisible by only itself and the unit [of measure].
asciilifeform: and this is only the beginning of the ugh
asciilifeform: mircea_popescu: if 1 gets to be a prime, you then throw out the conventional def of prime, because now you have a prime that divides other primes unequal to itself
spyked: mircea_popescu, yeha, but maths ppl I read/talked to "by convenience" define prime as "a number greater than 1 which is divisible ..." the reason escapes me. and as usual, online sources are shit.
mircea_popescu: 1 fits this, 0 does not.
mircea_popescu: spyked "a prime number is one which is divisible by only itself and the meter".
a111: Logged on 2017-09-20 19:36 mircea_popescu: anyway, the above example of "alt-addition" is mostly based on the happenstance that in base 10, the set of digits that are also prime is {1,2, 3, 5, 7} and the set of digits that are not prime is {0, 4, 6, 8, 9} ie it just so happens to be a perfectly balanced split.
spyked: http://btcbase.org/log/2017-09-20#1716177 <-- why 1? don't want to start a debate on that. dunno if debate was ever settled by mathematicians, but I learned in school that primes are strictly > 1 ☝︎
asciilifeform: well naturally whole thing is to be an article.
mircea_popescu: asciilifeform this works splendidly for blog article imo
asciilifeform: will be interesting if any reply, considering how the thing obsoletes half the idiocy they currently print
mircea_popescu: with the perhaps eventual result of actually producing a well grounded numeric-theoretic notion of what alf wanted, ie a fucking hash already
asciilifeform: mircea_popescu: i've been thinking about sending ffa , when finished, as an article into the saecular derps' 'cryptology journals', strictly for the lulz of getting their reject barf , and then posting, a la al schwartz
mircea_popescu: can come up with a few mp-extended additions of your own, and for that money even formulate a general theory of addition extension in this sense,
spyked: will have to work out a few more examples on paper.
spyked: oh. confusion is resolved then
mircea_popescu: in no case would the carry alter the current decimal level.
spyked rereads the definition of mp prime-addition
spyked: http://btcbase.org/log/2017-09-20#1716165 <-- I r confused. if 3 + 9 = 2 (+1 carry, because of primality of 3), then isn't the result 93? ☝︎
asciilifeform: but he dun have as many theorems.
mircea_popescu: i thought he was older
asciilifeform: asciilifeform's brother played it tho.
mircea_popescu: btw, i take it everyone has played knight's bounty ? as the crowning accomplishment of those eastern orc lands ?
asciilifeform: ( famously, kennedy bought entire railroad car right before signing the ban )
mircea_popescu: didn't obama win the cigar nobel prize ?
asciilifeform: for some reason i thought there were no ( at least Officially ) cuban cigars in usa
mircea_popescu: they're not expensive either, two for a benjie or somesuch
phf: asciilifeform: fyi they sell them within i suspect no more than half hour from you
asciilifeform: aluminum thing
asciilifeform: yea it is EXACTLY that shape
mircea_popescu: i suppose tube makes ok trainer dildo for 16yo debutante