Hide Idle (>14 d.) Chans
← 2017-10-13 | 2017-10-15 → ↓
BingoBoingo: !~ticker --market all
jhvh1: BingoBoingo: Bitstamp BTCUSD last: 5675.09, vol: 23439.06389521 | Bitfinex BTCUSD last: 5691.2, vol: 74451.50116852 | BTCChina BTCUSD last: 4280.016, vol: 0 | Kraken BTCUSD last: 5640.3, vol: 6894.7710041 | Volume-weighted last average: 5684.24724124
ben_vulpes: shinohai: do we have a from-zero "install gpg, ... now you're in the #trilema wot" ?
hanbot: ben_vulpes eulorum has one: http://www.eulorum.org/Account_Setup
apeloyee: http://btcbase.org/log/2017-10-13#1724915 << http://www.ada-auth.org/standards/rm12_w_tc1/html/RM-3-5-4.html , paragraph 21: "the range of Integer shall include the range –2**15+1 .. +2**15–1"
a111: Logged on 2017-10-13 15:17 asciilifeform: because 8 bits is not enough to index a useful number of 8-bit words
apeloyee: and paragraph 23: System.Max_Binary_Modulus shall be at least 2**16
a111: Logged on 2017-10-08 22:14 asciilifeform: http://btcbase.org/log/2017-10-08#1723058 << which op is it that only leaks 1 of its 2 operands ?
apeloyee: asciilifeform: do you plan to test how much information about operands leaks as difference in power consumption? It looks plausible that multiplying 0*0 and maxint*maxint consume different amount of power.
apeloyee: continuing elimination of large temporary variables, |E| in FZ_Mod_Exp needs to be replaced with an index into |Exponent|
apeloyee: it seems I see how to squeeze out one more FZ-sized temporary from FZ_Mod_Exp, besides the ^^ and using a proper barrett; it will exacerbate the above-mentioned physical leakage, though... http://p.bvulpes.com/pastes/XvDnd/?raw=true <- as usual, not tested.
BingoBoingo: !~bcstats
jhvh1: BingoBoingo: Current Blocks: 489789 | Current Difficulty: 1.123863285132E12 | Next Difficulty At Block: 489887 | Next Difficulty In: 98 blocks | Next Difficulty In About: 14 hours, 53 minutes, and 22 seconds | Next Difficulty Estimate: None | Estimated Percent Change: None
BingoBoingo: !~ticker --market all
jhvh1: BingoBoingo: Bitstamp BTCUSD last: 5688.0, vol: 14061.26071216 | Bitfinex BTCUSD last: 5734.8, vol: 52967.24614921 | BTCChina BTCUSD last: 4280.016, vol: 0 | Kraken BTCUSD last: 5695.5, vol: 4114.73692849 | Volume-weighted last average: 5723.27710452
apeloyee: http://btcbase.org/log/2017-10-07#1722378 << tried actually using mod-exp function, not just construcing mod-exp from multiplications and divisions?
a111: Logged on 2017-10-07 19:30 asciilifeform: the interesting imho discovery is that heathen bignumtrons don't win much (or even any!) speed by normalizing the ints being added/subtracted
mircea_popescu: http://www.chicagotribune.com/news/local/breaking/ct-chicago-violence-shootings-20171013-story.html << funny how it's "chicago violence" when da police execute random pedestrian.
asciilifeform: http://btcbase.org/log/2017-10-14#1725195 << we don't use signed type for anything
a111: Logged on 2017-10-14 09:36 apeloyee: http://btcbase.org/log/2017-10-13#1724915 << http://www.ada-auth.org/standards/rm12_w_tc1/html/RM-3-5-4.html , paragraph 21: "the range of Integer shall include the range –2**15+1 .. +2**15–1"
asciilifeform: http://btcbase.org/log/2017-10-14#1725197 << this is so. idea of quoted thread was, i'd like to get ffa to where it uses strictly 2 machine types, 'Word' (whatever that is on whatever machine) and, say, 'Index', which is guaranteed to be mod 2**16 or larger. and get rid of all conversions.
a111: Logged on 2017-10-14 09:40 apeloyee: and paragraph 23: System.Max_Binary_Modulus shall be at least 2**16
asciilifeform: http://btcbase.org/log/2017-10-14#1725200 << i strongly suspect that this is a waste of time on pc, where even add and xor almost certainly consume variant current depending on operands
a111: Logged on 2017-10-14 10:23 apeloyee: asciilifeform: do you plan to test how much information about operands leaks as difference in power consumption? It looks plausible that multiplying 0*0 and maxint*maxint consume different amount of power.
asciilifeform: i'm concerned with 1) timing leak 2) fits-in head --- strictly.
apeloyee: I read that as "Word is guaranteed to be at least 16-bit wide also"
mircea_popescu: moreover, power leak is reasonalby easily fixed.
asciilifeform: http://btcbase.org/log/2017-10-14#1725201 << i deliberately did not do this, because it adds loc and saves 0 time
a111: Logged on 2017-10-14 11:02 apeloyee: continuing elimination of large temporary variables, |E| in FZ_Mod_Exp needs to be replaced with an index into |Exponent|
asciilifeform: optimization that adds loc is for inner loop items STRICTLY
apeloyee: bbut fpga has little memory!!!1
asciilifeform: fpga has a bit muxer and none of the loops will be loops, lol
apeloyee: lol even the exponentiation loop?
a111: Logged on 2017-10-08 00:20 asciilifeform: http://btcbase.org/log/2017-10-07#1722411 << 1 ) ffa is closed form. i.e. it CAN be written as a number of nand gates, with a 'funnel' at the top, to which you present a,b,c, e.g. 4096bit, numbers, and at the bottom in a little cup you get a^b mod c , and with NO UPWARDS FEEDBACK FLOW of information , i.e. answer comes after same interval of time always, and with strictly downwards signals.
asciilifeform: FOR ALL OF IT
apeloyee: you don't have fpga that large.
asciilifeform: i'll have a square metre grid of whatever size i can buy, if i want.
apeloyee: i'll have a square metre grid of whatever size i can buy, if i want. << not enough wires connecting them.
asciilifeform: now this is nonsense.
shinohai: http://btcbase.org/log/2017-10-14#1725193 <<< hanbot has it, though I propose that when tmsr isp is functional we revisit idea of trilema wiki if mircea_popescu agrees.
a111: Logged on 2017-10-14 05:00 ben_vulpes: shinohai: do we have a from-zero "install gpg, ... now you're in the #trilema wot" ?
asciilifeform: http://btcbase.org/log/2017-10-14#1725202 << i will say now, that i will not be using windowed exponentiation, because it wins nothing unless you permit indexing memory by a secret
a111: Logged on 2017-10-14 12:06 apeloyee: it seems I see how to squeeze out one more FZ-sized temporary from FZ_Mod_Exp, besides the ^^ and using a proper barrett; it will exacerbate the above-mentioned physical leakage, though... http://p.bvulpes.com/pastes/XvDnd/?raw=true <- as usual, not tested.
asciilifeform: and i have no intention of doing such a thing
shinohai: I still have all the original articles saved from the #b-a days
apeloyee: for i in Base'Range loop B(i):=W_Mux(W_Mux(One(i), Base(i), WindowLSB),W_Mux(BaseModSquare(i),BaseModCube(i),WindowLSB),WindowMSB); end loop;
apeloyee: you get the idea.
mod6: mornin
mircea_popescu: shinohai i see no problem, what can it possibly hurt.
asciilifeform: apeloyee: actually no, i dun get ?
mircea_popescu: heya mod6
mod6: :]
shinohai: Buenas dias mod6
mod6 waves
asciilifeform: http://btcbase.org/log/2017-10-14#1725219 << in your house -- yes. in , say, a racked machine in a dc -- not so easy
a111: Logged on 2017-10-14 13:48 mircea_popescu: moreover, power leak is reasonalby easily fixed.
mircea_popescu: there is that.
mircea_popescu: of course, racked machine comes with keys preinstalled definitionally.
mircea_popescu: but yes, decryption still leaky.
mircea_popescu: see alf, had you listened to me when i FIRST told you to fucking move, you'd have had a warehouse somewhere outside of shit empire, could have just made your own dc.
asciilifeform: prolly the only finalsolution to power variance is http://btcbase.org/log/2017-05-16#1656782 .
a111: Logged on 2017-05-16 01:53 asciilifeform: now you store bit as ~phase~ of the square wave, rather than absolute steady state of the flipflop.
mircea_popescu: not sure that buys you anything, now you got an FM carrier wave in the power supply instead of a simple levels thing.
asciilifeform: you have a couple mil 'fm radios' less than a milliwavelength apart
mircea_popescu: mno i dun think so or mno i can prove it.
asciilifeform: (physically)
apeloyee: let's suppose B is a temporary, WindowLSB and WindowMSB are bits of 2-bit fixed window, One is, well, 1, BaseModSquare and BaseModCube are B**2 mod Modulus and B**3 mod Modulus. the posted snippet then assigns to B the power (2*WindowMSB+WindowLSB) of Base. See, we index the memory by a (small) secret!
mircea_popescu: lol this guy
mircea_popescu: apeloyee was your first sexual experience of the "just let me put the tip in" type ?
asciilifeform: apeloyee: nogood
asciilifeform: apeloyee: NO INDEXING BY SECRET OF ANY SIZE
asciilifeform: what part of this is hard to understand.
asciilifeform: no pasaran.
apeloyee: asciilifeform: where do you see explicit indexing in the posted snippet?
asciilifeform: nowhere, but not clear to me that it wins anything, and would have to be indexed this way in order to win
apeloyee: muliply-and-reduce is so expensive that 2-bit windows ought to win something.
asciilifeform: btw apeloyee what does this do when exponent is 1 ?
asciilifeform: 1 is not a sums of 2s and 3s...
apeloyee: 1 is not a sums of 2s and 3s... << ???
apeloyee: ah. I made an error.BaseModSquare and BaseModCube are Base**2 mod Modulus and Base**3 mod Modulus
apeloyee: not B of course
asciilifeform: it still isn't clear to me how this algo can modexp to the power of 1.
asciilifeform: and for that matter of 0.
mircea_popescu: special cases!
asciilifeform: lolfuckno
mircea_popescu won't rehash.
asciilifeform: no conditionaljumps on secret MEANS no conditionaljumps on secret. no exceptions.
apeloyee: no special cases. you precalc Base**0 = 1, Base**1 = Base, Base**2 and Base**3; assign 1 to result. then you take two next bits from exponent, starting with the most significant; take B to their power, as above, and multiply Result by B
mircea_popescu: what is your definition of special cases?
asciilifeform: apeloyee: and index with a 4way mux ? that'd work
asciilifeform: i still think that it makes sense to do this only after every other bolt is as tight as physically possible -- bernsteinian karatsuba, unrolled comba, etc
mircea_popescu: so every time you calc an exp you calc 2 and 3 also ?
apeloyee: so every time you calc an exp you calc 2 and 3 also ? << at the very start, once.
apeloyee: can has link describing " bernsteinian karatsuba"?
apeloyee: *link to description
asciilifeform: apeloyee: iirc it was in middle of https://cr.yp.to/ecdh/curve41417-20140706.pdf ( which is largely about other, useless crud )
asciilifeform: and there is literally no other description afaik anywhere on the net
apeloyee: every other bolt << does this include an asm multiply?
a111: Logged on 2017-08-10 02:43 asciilifeform: for simplicity, tested the case that actually happens in practice: on a 64bit box, any ffa width over 512 bits gives a strictly 8-wide comba mult ocurrence
asciilifeform: apeloyee: if doing asm, not only multiply but add and shift
asciilifeform: ( massive time is wasted deriving the carries )
asciilifeform: but i'd like an end product with NO arch dependence.
asciilifeform: i'ma eventually rewrite whole thing in asm, for special uses. but that will be separate proggy.
asciilifeform: canonical ffa will never contain asm.
asciilifeform: nor any endianisms or anything else typing it to particular iron.
apeloyee: procedure W_Mul is separate, etc. a generic version and an asm version. who dislikes asm, can throw it away and use generic version.
asciilifeform: understand what canonical means in this case ?
asciilifeform: it means the thing has to be gotten to acceptable performance sans asm.
asciilifeform: and the correctness of any variant (e.g. asmed) versions, will be judged against the canonical.
shinohai: "Hey Buddy, thanks for the tip!" http://archive.is/rlSnw
mircea_popescu: bwahahah
mircea_popescu: "Some of the emergency measures, notably beefed up police powers, are to be made permanent under a bill that has been approved by the lower house of parliament, the National Assembly."
mircea_popescu: this whole bruxelles thing has totally worked wonders at turning europe into an utter shithole.
mircea_popescu: moar "unity" and bullshit please, they totally work.
shinohai: http://archive.is/6bujY <<< But yeah, my account totally gets suspended for using a bot because pr0n.
mircea_popescu: lol pantsuit, they actually expect you can ban by reference.
mircea_popescu: hurr durr ban stupidity while at it.
mircea_popescu: three years late, but i WILL fucking sink this whole silicon valley social media bs.
shinohai: Don't open Slate this morning then, stupidity in form of "This Week, It Was Particularly Rough to Be a Woman" abounds.
shinohai: President Donald Trump raped me with his thoughts, etc.
mircea_popescu: working at making it rougher, actually.
ben_vulpes: hanbot: ty!
apeloyee: http://btcbase.org/log/2017-10-14#1725293 << lulzily, it mentions http://btcbase.org/log/2017-10-14#1725240 method on page 9: "we load the entire table into registers and perform the selection via arithmetic". wonder how you have read that paper if you didn't notice that.
a111: Logged on 2017-10-14 14:25 asciilifeform: apeloyee: iirc it was in middle of https://cr.yp.to/ecdh/curve41417-20140706.pdf ( which is largely about other, useless crud )
a111: Logged on 2017-10-14 13:55 apeloyee: for i in Base'Range loop B(i):=W_Mux(W_Mux(One(i), Base(i), WindowLSB),W_Mux(BaseModSquare(i),BaseModCube(i),WindowLSB),WindowMSB); end loop;
apeloyee: besides, "bernsteinan karatsuba" requres carry-save arithmetic, otherwise it likely wins nothing. so not separate from comba rewrite.
apeloyee: !#s from:asciilifeform carry save
a111: 3 results for "from:asciilifeform carry save", http://btcbase.org/log-search?q=from%3Aasciilifeform%20carry%20save
apeloyee: windowed mod-exp touches only mod-exp.
apeloyee: still, left-to-right exp (as inhttp://btcbase.org/log/2017-10-14#1725202 ) uses one FZ-sized temporary less than current ffa's right-to-left. (the indexing of E can be reverted to what ffa currently has).
a111: Logged on 2017-10-14 12:06 apeloyee: it seems I see how to squeeze out one more FZ-sized temporary from FZ_Mod_Exp, besides the ^^ and using a proper barrett; it will exacerbate the above-mentioned physical leakage, though... http://p.bvulpes.com/pastes/XvDnd/?raw=true <- as usual, not tested.
trinque: mircea_popescu: possible my network connection goes out again, but what's up?
mircea_popescu: trinque for to payments, 1 sec.
trinque: cool
mircea_popescu: !!withdraw 0.236986 1BoJ1wgM446w2yKWJrPjJD7sePvs13sagw
BingoBoingo: BWAHAHAHAHAHAHA, Eggspress
mircea_popescu: but you see ? if he can believe, as i not for a second doubt he does, that "a majority of websites are not built through coding", what exactly is to keep him from believing, and in the end being ready to die for, any other nonsense ?
mircea_popescu: "the majority of fortunes were not built by work". why not ? global warming, pantsuit.*
BingoBoingo: This is an incredibly common phenomenon
asciilifeform: http://btcbase.org/log/2017-10-14#1725324 << lol it is very very easy, when i see a paragraph that seems to be about ecc crypto, my eyeball jumps over it
a111: Logged on 2017-10-14 18:38 apeloyee: http://btcbase.org/log/2017-10-14#1725293 << lulzily, it mentions http://btcbase.org/log/2017-10-14#1725240 method on page 9: "we load the entire table into registers and perform the selection via arithmetic". wonder how you have read that paper if you didn't notice that.
asciilifeform: http://btcbase.org/log/2017-10-14#1725327 << i don't see how it needs carry-save
a111: Logged on 2017-10-14 18:39 apeloyee: besides, "bernsteinan karatsuba" requres carry-save arithmetic, otherwise it likely wins nothing. so not separate from comba rewrite.
a111: Logged on 2017-10-14 18:57 apeloyee: still, left-to-right exp (as inhttp://btcbase.org/log/2017-10-14#1725202 ) uses one FZ-sized temporary less than current ffa's right-to-left. (the indexing of E can be reverted to what ffa currently has).
asciilifeform: i'ma certainly try the fixedtimewindow thing at some point
asciilifeform: ( why not even bigger window, e.g 8 bits ? )
asciilifeform: http://btcbase.org/log/2017-10-14#1725340 << i will admit to being curious re what it is mircea_popescu thinks he can use those folx for. they have the apparent intellectual ability of a turnip.
a111: Logged on 2017-10-14 20:33 mircea_popescu: but you see ? if he can believe, as i not for a second doubt he does, that "a majority of websites are not built through coding", what exactly is to keep him from believing, and in the end being ready to die for, any other nonsense ?
mircea_popescu: asciilifeform i am using them for literal target practice, to teach a young kitten to hunt.
mircea_popescu: and she is learning admirably, and i am well satisfied.
mod6: nice
mircea_popescu: asciilifeform incidentally, do you recall the one with the kids making an "independent" chan ?
mircea_popescu: google has mysteriously forgotten it entirely, for any searches.
asciilifeform bbl : meat
mod6: !!up wyatt06
deedbot: wyatt06 voiced for 30 minutes.
deedbot: EF2246E285906AA9F6F55406D0F6314BA95B33EB registered as wyatt06.
mod6: nice :]
mircea_popescu: wyatt06 make sure you also register your freenode nick with chanserv.
wyatt06: hi i did that
mod6: yeah, we got his irc set up, so thats his nick now. got his key in, so that's good.
deedbot: http://trilema.com/2017/the-majority-of-things-are-build-through-eggpress/ << Trilema - The majority of things are build through eggpress
deedbot: http://trilema.com/2017/the-majority-of-things-are-build-through-eggpress/ << Trilema - The majority of things are build through eggpress
mircea_popescu: ah you're mod6 kid. welcome hey.
mircea_popescu: !!up alex__c
deedbot: alex__c voiced for 30 minutes.
alex__c: thanks
wyatt06: mircea_popescu: yup!
wyatt06: thanks
mircea_popescu: !!rate wyatt06 1 if he breaks anything invoice mod6
BingoBoingo: http://btcbase.org/log/2017-10-14#1725254 << In other developements on the belated addressing of old problems, It appears on of the small HK fish has referred inquires to bigger HK fish. More news if it goes anywhere.
a111: Logged on 2017-10-14 14:01 mircea_popescu: see alf, had you listened to me when i FIRST told you to fucking move, you'd have had a warehouse somewhere outside of shit empire, could have just made your own dc.
mod6: lol mircea_popescu
mod6: let's see if he can up himself here... standby
mircea_popescu: and in other lulz, loads of larger ops (including blockchain.info) are actually behind the chain tip now
mircea_popescu: !!withdraw 0.00953429 114KjtTEg1K17jgCRam7BqMjD8cCNYxdVu
mod6: dang
shinohai: dat ass
mircea_popescu: in continuing lulz, "All my ETH is now invested in alts".
lobbes: buterin's waterfall keeps a-flowin'
← 2017-10-13 | 2017-10-15 → ↑
Random(trilema) | Download hourly DB snapshot | Get Source Code