asciilifeform: wb Framedragger !

asciilifeform: aaah gaijin

asciilifeform: BingoBoingo: what's a barang ?

asciilifeform: d00d evidently did his homework.

asciilifeform: in other lulz, http://archive.is/wskHt ☟︎

asciilifeform: nuts.

asciilifeform: this piece was memorable, because at no point was meat any kind of luxury for asciilifeform & co.

asciilifeform: did i dream this?

asciilifeform: where girls in mircea_popescu's school would beg him for meat or how did it go

asciilifeform: alluded to in http://btcbase.org/log/2017-02-19#1615576 but iirc was in a trilema piece ☝︎

asciilifeform: hm could've sworn it was in l0gz

asciilifeform: interesting combo, boys fighting over meat in soup, but left-handing -- tolerated

asciilifeform: oh hm.

asciilifeform: vlad's dukedom iirc ?

asciilifeform genuinely surprised

asciilifeform: mircea_popescu: i take it this custom was not alive in classical ro ?

asciilifeform: aha.

asciilifeform: incidentally ( and luckily for asciilifeform ) asciilifeform is not left-handed : left-handers were savagely beaten on asciilifeform's home world

asciilifeform currently making an efficient seekritshift, it is somewhat tricky

asciilifeform: aaa

asciilifeform: waiwat who

asciilifeform: !#s garza

asciilifeform: tldr: usg inquisitor imposes 10mil $ fine, plus whatever jail ( decided later )

asciilifeform: !~later tell mircea_popescu http://wotpaste.cascadianhacker.com/pastes/93H3s/?raw=true << possibly lulzy: garza convicted

asciilifeform bbl.

asciilifeform: ok laters then

asciilifeform: and on what does their number depend.

asciilifeform: but what are the fixed shots ?

asciilifeform: yes that's one shot

asciilifeform: can i persuade you to write this out as algo , apeloyee ?

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: pretty sure you just described lehmer's gcd, actually

asciilifeform: thing has to take exactly.same.number.of.steps. regardless of what a,b equal.

asciilifeform: that ain't constant time !!

asciilifeform: q remains tho, how to handle a==b case

asciilifeform: tru!!

asciilifeform: time will vary massively depending on hamming weights

asciilifeform: other thing is that i dun see how this is constant time

asciilifeform: what then ?

asciilifeform: and say they're equal.

asciilifeform: so i sorted them, answer is for the sorted pair then, not the actual pair neh

asciilifeform: what happens if a <= b ??

asciilifeform: let's have it

asciilifeform: oh??

asciilifeform: apeloyee: if you can think of a subquadratic constanttime gcd, plz do write in

asciilifeform: hm, i must've been thinking of some other gcd algo, because lehmer's is a definite dead end.

asciilifeform: ( or not ? )

asciilifeform: currently i suspect that it is possible to constantize lehmer's logn gcd.

asciilifeform: ( and still not constant time )

asciilifeform: euclidean'd be o(n^3) yes ☟︎

asciilifeform: i dun have a working gcd yet

asciilifeform: take lehmer's gcd as a starting point.

asciilifeform: why do you think constanttime gotta be, apeloyee ?

asciilifeform: non-constanttime gcd is certainly not O(N^3)

asciilifeform: trial division is dog slow vs gcd. ☟︎

asciilifeform: and in who knows what future numbertheoretical ops. we gotta have gcd.

asciilifeform: the other obvious place for gcd is rsa phi

asciilifeform: save massive time

asciilifeform: want to gcd(candidate, biggestprimorialthatfitsintheffabitness) ☟︎☟︎

asciilifeform: for the initial sieve ~prior~ to miller-rabin ☟︎☟︎

asciilifeform: miller-rabin

asciilifeform: eagle eyes, apeloyee . i dun suppose you have a constant time gcd up your sleeve ?

asciilifeform: !!v 17BC89B25DD39E071B293EDE353D50C10A165F320E436ED7FE36D996140DBBCC

asciilifeform: !!rate apeloyee 2 A - N*floor(A*R/4^K) < 2*N

asciilifeform: !!up apeloyee

asciilifeform: ok indeed worx.

asciilifeform: ok

asciilifeform: let's start with the above

asciilifeform: dunno if i do, still trying to swallow the proof.

asciilifeform: i can read, and see this. but why is it 1 ?

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 )

asciilifeform: neato apeloyee , ty.

asciilifeform: adds exactly one sub-and-mux to the existing 3.

asciilifeform: oh hm this worx.

asciilifeform: apeloyee: so what'd the correction step look like, with the pseudoquotient

asciilifeform: aa 2^(2k) ok

asciilifeform: why 4^k ?

asciilifeform: apeloyee: what is B ?

asciilifeform rereads

asciilifeform: i was discussing a 'shiftless' doublewide barrett

asciilifeform: as i understand will be quite exactly it.

asciilifeform: apeloyee: it won't differ at all, once i have working shifter using your scheme.

asciilifeform: apparently ~nobody else properly reads.

asciilifeform: ty for properly rubbing it in asciilifeform's nose, apeloyee

asciilifeform: it gotta

asciilifeform: aha

asciilifeform: same thing we do for the wholeword shift

asciilifeform: ( first : 'maybe shift by 1' , second : 'maybe by 2', last : ' ... by 32' )

asciilifeform: ( example : if wordbitness is 64, any subword shift is from 0 .. 63, and therefore can be expressed using 6 muxgated shifts . )

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: !!up apeloyee

asciilifeform: ftr apeloyee has entirely valid, imho, algo, but terribly confusing pseudocode

asciilifeform: and possibly he will agree to swap it in.

asciilifeform: apeloyee: please make a new one ? and give to trinque , signed with old

asciilifeform: http://wotpaste.cascadianhacker.com/pastes/D8m5z/?raw=true << gory detail

asciilifeform: subkey, to be specific.

asciilifeform: why didja make a key that expired in weeks, apeloyee ?!

asciilifeform: ( in trinque's planet, but not yet here )

asciilifeform: and key expired.

asciilifeform: trinque's clock is ahead of mine