mod6 sprays down the whiteboard
a111: Logged on 2017-05-21 16:47 asciilifeform: because ALL ops take same time, so karatsuma, toom-cook, etc. cannot work because they fundamentally rely on breaking large x*y into a number of smaller a1*b1, a2*b2, ...
mod6: ah, and there was a time when L was not part of the plan
mod6: mircea_popescu: my eyyyes
mod6: asciilifeform: indeed, was just re-reading. makes sense.
mod6: f, my ffa is super old
mod6: nevermind found one
mod6: asciilifeform: you mind throwing me your W_Add_D, i don't seem to have one.
mod6: are 'Lo : Word; \n Hi : Word;' initialized to zero on the first pass through Col(...) by default?
mod6: ok, my bad, i see those are the output words for W_Mul, XY_LW and XY_HI.
mod6: (im going through by hand here...)
mod6: lol, im gonna need a bigger board.
mod6: lol, im probably just dim, but i think it's the only way im gonna understand things thing through and through
mod6: my gurl is like X_X
mod6: workin through this on the board. so!
mod6: i did solve a few of my own questions so far on it.
mod6: and gleaned a few things I didn't notice earlier.
mod6: i haven't gone all the way through the W_Mul yet, but just unrolling it as said.
mod6: for i in L .. 2*L - 2 loop
mod6: Col(i, i - L + 1, L - 1);
mod6: this was throwing me off earlier today ^
mod6: but now, it's starting to make sense.
mod6: now that i've got that part in my head, now can tackle the rest tomorrow.
mod6: damn, i wanna be on whatever deck she's on.
ben_vulpes: self-administered "remedial binary arthmetic with practical applications" course is very educational; going well
☟︎ mod6: ok neat, i just ran it against my V stuff, seems to dump out all of the symbols or whatnot
mod6: ah, that's sweet asciilifeform
mod6 reads the rms thread
deedbot: fromdeedbot voiced for 30 minutes.
BingoBoingo: In other news: Fake news attacking idiot son (Donald Jr.) for trying to impress father because shut out of campaign as an idiot
shinohai: fake news is where the $ is at these dayz
a111: Logged on 2017-07-11 16:30 ben_vulpes: self-administered "remedial binary arthmetic with practical applications" course is very educational; going well
mircea_popescu: "As recently as earlier today, the Fake News empire turned idiot son Donald Junior's gullibility with respect to a phishing scheme1 into their latest headline of the day which they will use to sell loyal consumers of their fiction product on the "certain end" of the 8 year Trumpreich in spite of all the actual evidence running to the contrary of their narrative supported instead by a complete lack of evidence."
ben_vulpes: mircea_popescu: slowlybut surely, expanding the horizon of things i understand, taking the time to make sure i understand each part.
ben_vulpes: byte arrays to integers for the m+1, m+2, coercions for byte arrays to integers, bit arrays for flipping random bits,
mircea_popescu: karpeles and carl force "trials", parallel lulz constructions.
mircea_popescu: ben_vulpes here's a high pay grade question for you : of the two models of "controlled de-entropy" i spawned in a week, specifically a) count of randomly placed flipped bits, as in the discussion with you re that and b) string of randomly initiated, n bit long SET bits, as discussed in
http://btcbase.org/log/2017-07-10#1681268 which does the bitcoin foundation regard as a better candidate for standardization as "the republic'
☝︎ a111: Logged on 2017-07-10 19:50 mircea_popescu: which incidentally brings us to a very workable and very useful tmsr definition of entropy quality : take a FG string. flip a number of consecutive bits to 1. the result is your entropy quality, such as 100/1mb if you flipped 100 bits.
mircea_popescu: let M = FG[1Mb] ; let P = rnd(0, 1Mb) ; let M[P..P+100] = 1.
mircea_popescu: produces a bit-addressable array of 1048576 bits from FG
mircea_popescu: M' is now a "known low entropy bitfield". we know it to have degraded by 100/1Mb or w/e the case may be
mircea_popescu: then we feed it into entropy-dependent processes (say the rabin miller test, as discussed yest) and see what comes out.
mircea_popescu: in a sense. it has multiple utilities, it allows you to try and guess (numerically) whether for instance better entropy or more passes are useful for rabin miller
mircea_popescu: which allows you to correctly balance your tester irl.
mircea_popescu: as in, balance it according to a calculation, rather than according t oa feeling.
mircea_popescu: asciilifeform there is that, yes. but we're doiong this for instrumentation in teh lab not for any other purpose. it's a tracer for entropy, like the shit they make you swallow to see your stomach.
BingoBoingo: <mircea_popescu> wtf, newswriting not dictionary boating accident << Cleanup ongoing. Probably ought to write before baking brain in heat or sleep and proofread before publishing. /me resets "Days since linguistic boating accident" sign back to 0 days
shinohai likes mircea_popescu 's barium enema analogy
mircea_popescu: poor ben_vulpes has been silent since the blessing of wtf 10 minutes ago.
mircea_popescu: (ie, i dun have an answer for it myself, nor know how to get one.)
mircea_popescu: as alf points out, a is "better" in the sense v-n debiaser kills b, if this is a better. unclear.
mircea_popescu: a is also twice as strong, because obviously b will flip on average half the bits it sets. but is this ACTUALLY half ? ie, how do yo umeasure unrandomness ? b is much more strongly patterned than a.
mircea_popescu: asciilifeform no that's exactly what we're doing. controlledly-bad-level entropy.
mircea_popescu: the whole discussion is, given a stream of perfect entropy, how to construct known-degree-of-badness out of it.
mircea_popescu: and in the process see also that ent is not a very good measurement tool.
mircea_popescu: pass-fail tools notoriously horrendous measurement instruments.
mircea_popescu: asciilifeform understand this isn't a matter of testing specified bitfields, but specified ~processes~
shinohai: Would be a lulzy Qntra BingoBoingo
mircea_popescu: next the usg is going to get off its ass, discover the whole rapereum "ecosystem" was $7600.
mircea_popescu: gate, switches on at some point pisses 100 1s, switches back