Forum logs for 20 Jan 2019

Monday, 16 March, Year 12 d.Tr. | Author:
feedbot: http://qntra.net/2019/01/increased-usg-psyops-against-maduro-government-intensify-risk-of-military-confrontation/ << Qntra -- Increased USG Psyops Against Maduro Government Intensify Risk Of Military Confrontation [00:06]
BingoBoingo: ^ Yes I raised the alarm. Efforts to convince me to lower it are welcome. If you are outside my L1 you need to offer a stake and odds for your petition to be considered. [00:12]
BingoBoingo: https://twitter.com/BBoingo/status/1086854906055741440 << mircea_popescu Seems I have yet to get twatter banned [00:23]
mircea_popescu: apologetic girl -- "today's the last of my hormone pills, might have something to do with it". deaf mp -- "whore moan pills ?!" [02:36]
mircea_popescu: meanwhile in antiqua, http://btcbase.org/log/2013-12-21#428326 [10:29]
a111: Logged on 2013-12-21 02:07 asciilifeform: http://www.usatoday.com/story/nation/2013/12/19/maj-gen-michael-carey-nuclear-missiles-alcohol/4131963/ [10:29]
feedbot: http://bvt-trace.net/2019/01/experiment-n-way-split-karatsuba/ << bvt's backtrace -- Experiment: N-Way Split Karatsuba [10:31]
mircea_popescu: ahaha check him out asciilifeform ! MATHEMATICAL NOTATION! [10:35]
mircea_popescu: apparently, the split variant is ~slower~. [10:36]
bvt: hello [10:42]
bvt: re math, i found that htmling it would be unreadable, so decided to use images. [10:43]
bvt: as far as code speed is concerned, i still don't know whether i fucked up something, or the algorithm is fundamentally slow after some bignum size [10:44]
bvt: asciilifeform: http://bvt-trace.net/vpatches/ffa_ch1_genesis.kv.vpatch.bvt.sig http://bvt-trace.net/vpatches/ffa_ch2_logicals.kv.vpatch.bvt.sig http://bvt-trace.net/vpatches/ffa_ch3_shifts.kv.vpatch.bvt.sig [10:44]
bvt: http://bvt-trace.net/vpatches/ffa_ch4_ffacalc.kv.vpatch.bvt.sig http://bvt-trace.net/vpatches/ffa_ch5_egypt.kv.vpatch.bvt.sig http://bvt-trace.net/vpatches/ffa_ch6_simplest_rsa.kv.vpatch.bvt.sig [10:45]
bvt: will provide more seals after i work with chapters 7-9 more. [10:46]
mircea_popescu: bvt if you think about it, it has to be slower, because operations are fast and allocations slow. [10:49]
asciilifeform: bvt, mircea_popescu : http://bvt-trace.net/2019/01/experiment-n-way-split-karatsuba/comment-page-1/#comment-6 [10:55]
bvt: yes, and it's also true for memory accesses. however the number of executed instructions also increased a lot, which makes me suspect that there is something i missed [10:55]
mircea_popescu: i don't think so. [10:56]
asciilifeform: bvt: ty for reading, signing, and publishing experiment -- i will include your seals in ch16 article [10:57]
mircea_popescu: i suppose you might say "it shouldn't be x2, my expectation would be it's x2^1/2", but w/e. [10:57]
asciilifeform: mircea_popescu: i'm not surprised, considering that bounds check overhead (in ada with all safeties switched on) magnifies the 'losing' of a losing (for particular width) algo [10:58]
bvt: i would not be surprised it was 20% slower, but 2x was surprising for me [10:59]
bvt: asciilifeform: yw [10:59]
mircea_popescu: asciilifeform aha. kinda my thought also. [10:59]
asciilifeform: bvt: you have more than twice the # of cache evictions [10:59]
asciilifeform: of course 2x slower. [10:59]
asciilifeform: 'locality' effect is actually why e.g. comba wins over ordinary 'kindergarten' o(n^2) mult [11:00]
asciilifeform: despite both being in same order [11:01]
asciilifeform: same effect applies to karatsuba, and errywhere else. [11:01]
bvt: also, the linked pdf contains one FFT-like algorithm i considered to implement (using walsh transform for convolution instead of fft). but it'd be even more complex [11:01]
asciilifeform: i saw. and imho would be interesting to have a constant-spacetime, no-floats fft [11:02]
asciilifeform: ( not particularly useful for ffa, but potentially elsewhere.. ) [11:02]
bvt: yes, actually in this karatsuba/toom algo one can embed comba's trick, but the code would become even gnarlier [11:03]
bvt: re fft: i would wait until a use-case appears, to at least understand what are the requirements. do you have something in mind? [11:04]
asciilifeform: bvt: i don't, which is why never bothered with fft [11:07]
asciilifeform: i assumed bvt knew of a use, given that he dug out the ru materials on subj [11:07]
bvt: nope, but there were a few mentions of other algorithms in the logs, so i decided to have a look. don't remember how i arrived at this particular one [11:09]
mircea_popescu: lol, you're supposed to remember! [11:09]
asciilifeform: 'i built atomic dirigible but fughet why!111' [11:09]
asciilifeform: http://btcbase.org/log/2019-01-20#1888467 << gotta nitpick here: it aint allocations (which in ffa planet are always done by stack frame, in O(1) ) that leads to slow, but cache eviction ( as well as linear overhead from doing moar ops in general ) [11:11]
a111: Logged on 2019-01-20 15:49 mircea_popescu: bvt if you think about it, it has to be slower, because operations are fast and allocations slow. [11:11]
bvt: no, i mean that particular paper. i initially wanted to implement a fft-like multiplication algorithm, but got interested in this karatsuba/toom when reading the paper. [11:11]
mircea_popescu: nickpit granted! [11:12]
bvt: :-) [11:12]
mircea_popescu: anyway, nothing wrong with it, i enjoyed reading. [11:13]
mircea_popescu: there's that old joke re "i read good books twice but bad books i don't read at all" which very much applies : if only one knew before looking whether an algo is fast or slow! [11:13]
asciilifeform: there was another fella (since wandered off) who posted last yr an attempt at de-recursivizing karatsuba ( it's definitely in the logs, can't find it just yet tho ) [11:14]
bvt: mircea_popescu: thanks [11:15]
asciilifeform: bvt: there's a long list of things that asciilifeform considered and (for time being) rejected from ffa, on acct of costing substantial complexity for very small saving of cpu cost. e.g. unrolled comba. [11:16]
asciilifeform: sorta why, when i first started preparing the thing for publication, wound it back to the simplest known ('egyptian') variant, and walked from there. [11:17]
asciilifeform: it allows reader to see what is bought by adding the moving parts. [11:17]
bvt: asciilifeform: a better attempt at this algorithm can involve using comba at the lower level. [11:17]
bvt: myeah, complexity does not go too well with both constant-time and fits-in-head. [11:19]
asciilifeform: bvt: comba itself ended up on the list of things that made it in by very small margin ( it wins perhaps 10%, on most pc iron , vs straight word*word mul as base case, try it yourself by turning the threshhold knob ) [11:20]
bvt: will do. so far i had a very brief look at it [11:21]
asciilifeform: as i noted previously -- i do not expect to find any moar ~asymptotic~ speedups for ffa algos , such that are relevant to the sizes of numbers typically used in public key crypto [11:23]
asciilifeform: the remaining virgin land for speed revvup, is asmism for base cases. ( and ~possibly~ in combination with unrolled comba ) [11:24]
asciilifeform: http://www.loper-os.org/pub/ffa/hypertext/ch15/w_mul__adb.htm#95_14 << specifically here, currently we do 4 MUL instrs for where really needs only 1 [11:26]
bvt: my understanding is that asmism would go only for lower-level ffa code, i.e. barret/modexp will remain as-is. [11:26]
asciilifeform: ( and when ADD, SUB, we do it twice, where only needs once ) [11:26]
asciilifeform: bvt: correct [11:26]
asciilifeform: the only reason asmism even potentially invites itself, is that idjit compiler gives no primitive for add/sub-with-carry or full-word mul [11:27]
asciilifeform: the other 'secret' speed boost is if one were to turn off the bounds checks this gives ~2x speedup across the board. but i dun expect to do this for my personal uses for years , if ever -- it requires 100% certainty of correctness of the program under all possible input [11:28]
asciilifeform: ( and even presuming 100% correctness, bounds checks still give added 'cosmic ray resistance'. i ~like~ having'em ) [11:29]
bvt: http://p.bvulpes.com/pastes/q0ffh/?raw=true - some things can be done with gcc specific features, but i guess asming is cleaner [11:30]
asciilifeform: the ~other~, tho less simple, speed boost, is if one were to obtain a machine with wider multiplier ! [11:30]
asciilifeform: but i dun currently have such a thing. [11:30]
asciilifeform: bvt: i'd prefer asmisms to c imports ( which not only ugly and compiler-dependent, but i suspect destroy performance with overhead ) [11:31]
asciilifeform: recent pc irons offer 128x128bit iron mul. but i have not investigated it, and it specifically gotta be tested for constancy of time [11:32]
asciilifeform: in a hypothetical asmistic branch of ffa, you'd want to implement whole comba in asm, rather than merely word mul [11:34]
asciilifeform: same goes for add/sub. [11:34]
asciilifeform: ( otherwise you end up eating the overhead from the existing portable carry calculation mechanism for no reason ) [11:35]
mircea_popescu: i suspect double-wide mul might be implemented by lookup tables. [11:37]
asciilifeform: mircea_popescu: expand ? [11:38]
bvt: i.e. W_Borrow/W_Carry still cause the overhead? [11:38]
asciilifeform: bvt: correct [11:38]
bvt: i'll be afk an hour or two [11:39]
mircea_popescu: asciilifeform ie, splits the ints in bus width chunks and only does all 4 quadrants depending on results from the others. [11:39]
asciilifeform: mircea_popescu: that wouldn't be constant-time... [11:40]
mircea_popescu: which is my point. [11:40]
mircea_popescu: 128*128 bit iron mul is not ct by itself. [11:40]
asciilifeform: ( not to mention, a, e.g., 64 bit multiplication table, with extant logic density, would be approx the size of solar system ) [11:40]
asciilifeform: mircea_popescu: it is conceivable that the ones currently sold are constant time , i simply haven't tried'em. [11:41]
mircea_popescu: nah, just, if you're multipling (for simplicity) 01 x 01, it won't do 0x1 0x0 0x1 parts [11:41]
asciilifeform: ( the 64x64 iron multer in amd/intel, possibly surprisingly, is in fact constant time, in all boxes i've tested to date ) [11:41]
mircea_popescu: cuz cheap enough. [11:41]
asciilifeform: the # of gates req'd, is a cube. [11:42]
mircea_popescu: but afaik the double-wide mul, "optimized" [11:42]
asciilifeform: mircea_popescu: possibly, it'll have to be tested when asciilifeform or somebody else can be arsed [11:42]
asciilifeform: note that the 'cube' observation only applies if you're going for a single-clock-cycle iron multer. otherwise it grows as square of bitness. [11:43]
asciilifeform: ( the gate count, that is ) [11:43]
* asciilifeform has quite thick binder of curated material on subj, for the hypothetical day that we start baking irons [11:44]
asciilifeform: 1 annoying aspect of 'iron ffa'-gedankenexperiment, is that none of the available fpga ( either 'ice40' series, or the evil ones ) are anywhere near big enuff to prototype with. it'd have to be simulated a la http://www.loper-os.org/?p=2593 , slowly, and then straight to silicon. [11:46]
* asciilifeform brb,teatime [11:47]
mircea_popescu: meanwhile in news : http://btcbase.org/log/2019-01-19#1888338 << mystery solved. trb uses bdb for block indexing the index had an incorrect lobe affecting some blocks whenever someone asked for one of those blocks, bdb died and trb never returned. [12:17]
a111: Logged on 2019-01-19 13:50 mircea_popescu: aaand in wtf lulz, http://btcbase.org/log/2019-01-14#1886738 item died again, same exact details, same exact item, jan 15th (ie, it ran for ~day). ima finally get my ass into gear and debug-run it. [12:17]
mircea_popescu: -reindex rebuilds the index as expected. fin. [12:18]
mircea_popescu: meanwhile in trilema lulz : netstat | grep http | wc -l > 190. [12:20]
mircea_popescu: apparently 200 reqs ~per 100ms~ (that's how long it takes to complete one) is not enough to bring it down. [12:20]
asciilifeform: mircea_popescu: lobe << that's still interesting - invites q of ~why~ had bad lobe [13:15]
asciilifeform: ... disk rot ? [13:15]
asciilifeform: i prolly oughta add to the http://btcbase.org/log/2019-01-20#1888508 thing : 1 of the items which seemed like a speedup, but in actual practice sucked, was the use of (constant-time) 2 (ditto 4) -bit windows for modexp ( iirc apeloyee suggested ) [13:47]
a111: Logged on 2019-01-20 16:23 asciilifeform: as i noted previously -- i do not expect to find any moar ~asymptotic~ speedups for ffa algos , such that are relevant to the sizes of numbers typically used in public key crypto [13:47]
asciilifeform: the overhead eats the winnings. [13:48]
asciilifeform: possibly i'ma do a writeup on the subj, once errything else is fielded. [13:48]
asciilifeform: ( in all fairness, a large -- e.g. 8bit -- window, ~could~ win, but massively multiplies the memory requirement for the thing ) [13:52]
asciilifeform: chances are that it wouldn't, tho, given how the table still has to be indexed via fz_mux in order to prevent variant (i.e. nonconstanttime) memory indexing [13:53]
asciilifeform: ( '1st commandment' of ffa : thou shalt not branch on seekrit bits. '2nd commandment' -- thou shalt not index memory by seekrit bits ... ) [13:54]
asciilifeform: btw, must also add to bvt's http://btcbase.org/log/2019-01-20#1888517 >> https://gcc.gnu.org/onlinedocs/gcc/Integer-Overflow-Builtins.html admits that : 'The compiler will attempt to use hardware instructions to implement these built-in functions where possible, like conditional jump on overflow after addition, conditional jump on carry etc.' , i.e. there is not a guarantee that the thing dun introduce cond jumps. [13:58]
a111: Logged on 2019-01-20 16:30 bvt: http://p.bvulpes.com/pastes/q0ffh/?raw=true - some things can be done with gcc specific features, but i guess asming is cleaner [13:58]
asciilifeform: therefore that approach is completely verboten. [13:59]
asciilifeform: a correct asmism would simply read the carry flag and put the value where it belongs (e.g. in fz_add, into the next addition, in comba -- into the accumulator etc) [14:00]
asciilifeform: without any jumps [14:00]
asciilifeform: gcc's knob seems to be geared for scenarios where the overflow is an error condition, rather than expected. [14:01]
bvt: agreed. on x86_64 is compiled to ADD SETC sequence, but who knows what happens on other arch/gcc version. [14:02]
asciilifeform: ( ada standard btw trivially allows for types where this holds true automatically , i.e. throws exception for overflow. but this is not only massively unconstanttime but the overhead is gigantic ) [14:02]
asciilifeform: bvt: 1 of the reasons why ada doesn't offer e.g. addition-with-overflow , is that there is an abundance of sad iron where there isn't even physically a carry flag. [14:03]
asciilifeform: misguided folx ~continue~ to build these, with the excuse given being 'pipeline' [14:04]
asciilifeform: e [14:04]
asciilifeform: .g., [14:04]
asciilifeform: !#s riscv [14:04]
a111: 11 results for "riscv", http://btcbase.org/log-search?q=riscv [14:04]
asciilifeform: i find it interesting that -- afaik -- nobody's ever built iron that was specifically optimized for bignum [14:05]
asciilifeform: and no it dun have to have gigantic bus width, necessarily [14:05]
bvt: i wonder how this is a valid argument even there. if nothing reads the flag, there are no pipeline dependecies. [14:06]
asciilifeform: could simply have optimized 'take these-here N words and those-there M words, and put bignum addition in memory starting at O, and overflow flag in P ' or similarly [14:06]
asciilifeform: bvt: notion was, you gotta stop the pipe if something ~were~ to read it [14:06]
bvt: perhaps some old APL machines would qualify [14:06]
asciilifeform: bvt: possibly the bolix machines also ( they did it in vertical microcode, iirc, tho, and in nonconstant time unsurprisingly ) [14:07]
bvt: actually, that fft-like multiplication algorithm was developed in APL [14:08]
asciilifeform: in simple o(n) bignum operations like addition, the cost of instruction decoding for each consecutive 'add' , is substantial % of the cost [14:09]
asciilifeform: so even vertical microcode would win [14:09]
bvt: https://www.dyalog.com/uploads/conference/dyalog12/presentations/U23_MultidigitAlgorithms/Optimization%20of%20parallel%20multi-digit%20algorithms.pdf [14:09]
asciilifeform: bvt: i found a buncha these, when digging [14:10]
asciilifeform: often called 'systolic' algos [14:10]
asciilifeform: they potentially win, but only on custom iron really [14:11]
bvt: are such machines even build these days? i.e. for dsp, or other use-cases? [14:12]
asciilifeform: bvt: there's a large market of various voodoo for dsp, but afaik it all operates 'inexactly' [14:13]
asciilifeform: ( and defo not in constant time ) [14:13]
asciilifeform: more or less entirely opposite approach from what's wanted for crypto. [14:14]
asciilifeform: ftr i suspect that entirely ordinary algos, such as are seen in the current ffa, would already give ~line-rate~ (i.e. , 4096 modexp faster than 1G/s nic can give you new inputs to modexp on ) if implemented in iron properly. [14:21]
asciilifeform: without any exotic systolicisms etc [14:22]
asciilifeform: helps to recall that the problem which originally prompted asciilifeform to write ffa, is a (currently hypothetical) application where rsa sigs are carried in ~individual packets~ [14:23]
asciilifeform: ( is why all previously published rsatrons , entirely unsuitable -- if there's any leakage at all via timing, enemy trivially derives yer key ) [14:24]
asciilifeform: nao, it isn't as if the current ffa, with 2.7sec 4096-bit modexp, is immediately usable to eat packets at line rate. but that part at least theoretically parallelizes ( i.e. a rack fulla multicore boxen running ffa, can theoretically eat packets at line rate... ) [14:26]
asciilifeform: err, 1.7 [14:27]
asciilifeform: 1G/s link in principle delivers 262144 4096bit packets /sec ( in practice, many fewer, on acct of overhead ) [14:28]
asciilifeform: anyway pretty sure we had this thread, is in the logs somewhere. [14:28]
asciilifeform: before considering to bake irons, it is worth to see what a 100%-asmic ffa would give. [14:30]
asciilifeform: ( this is not a contradiction in terms, it is possible to implement whole thing, with same constant-time algos, by hand asm ) [14:30]
asciilifeform: ideally one would unroll ~all~ of the loops ( e.g. instead of looping through the words of a bignum, would e.g. add-with-carry on immediately consecutive words with stream of add instrs, etc ) [14:32]
bvt: would this actually be efficient on current arches? this would stress instruction decoder a lot [14:34]
asciilifeform: bvt: i expect one would trivially get a 10-20x speedup over the ordinary ffa, esp. if the item still fits in l1 [14:35]
asciilifeform: i found last yr, for instance, that unrolled comba ( still in ada ) gives 20-25% speedup. [14:36]
asciilifeform: the penalty from having any branches, of whatever kind, anywhere at all, on pc iron -- is substantial [14:37]
asciilifeform: ( consider the http://www.loper-os.org/?p=2906#selection-864.0-883.127 experiment, for instance. ) [14:38]
asciilifeform: koch's turd, despite being implemented in c, with no bounds checks, actually loses to ch14 ffa , for inputs of same ~width~ -- despite fact that he doesn't constanttime and thereby gets to skip massive work [14:39]
asciilifeform: the skipping itself is expensive enuff on iron with cache/branchpredictor, that he loses rather than wins from it. [14:39]
bvt: i guess i could do some experiments here. the immediate question is that ffa does plenty of FZ_Adds with different FZ'Length, so full unrolling would not really work (unless i miss something). [14:41]
asciilifeform: bvt: all of the lengths are deterministically known from the primary fz width. [14:41]
asciilifeform: therefore you can unroll. [14:41]
asciilifeform: just as you can de-recursivize the karatsuba etc [14:41]
asciilifeform: none of the lengths depend on the actual contents of the user input. [14:42]
asciilifeform: granted, an unrolled ffa would operate on a fixed width (e.g. 8192) of primary fz. [14:42]
asciilifeform: but this limitation would also be true of any hypothetical arithmetic iron. [14:43]
bvt: aha, i see. this would also involve lots function call inlining as well [14:44]
asciilifeform: bvt: even the current (ch11 and after) ffa relies on a gnat with working forced-inlining [14:44]
asciilifeform: ( ave1 discovered how to guarantee working inlining, and this gave 'free' ~2x speedup ) [14:45]
bvt: writing such code really could use some program to generate the necessary amount of adds/subs [14:45]
asciilifeform: bvt: a typical macroassembler would work for the purpose. [14:45]
asciilifeform: one would still want to audit the output of any such thing tho, by hand. [14:45]
bvt: ave1's fix also fixed the 'undefined references' in static lib with gcc-6.3 for me [14:46]
asciilifeform: not having used gcc5+ , i never saw this bug [14:46]
bvt: you can't verify output of 'write 4096 adds' by hand, though [14:47]
asciilifeform: bvt: why not ? they all will look same neh [14:47]
asciilifeform: it's no moar complicated than to count bricks in a wall. [14:47]
asciilifeform: and besides, the wall wouldn't even have that many bricks -- on a 64bit bus, a 4096-bit addition is 64 instructions long [14:48]
asciilifeform: ( if you have an add-with-carry instr ) [14:48]
asciilifeform: there's an obvious limit to what you can unroll and still fit in any plausible cache tho. [14:49]
asciilifeform: ( you won't be unrolling an entire 4096bit modexp, on any plausible irons... ) [14:50]
bvt: this is true, rigth. [14:50]
asciilifeform: the inner loop or 2, tho, definitely can. [14:50]
asciilifeform: i'll add , for completeness of thread, that if yer ~sending~, rather than receiving, rsa packets, your bottleneck will be ~rng~ long before it could ever be the arithmetron per se [14:55]
asciilifeform: ( a single FG , recall, yields ~7kB/s at room temp ) [14:56]
asciilifeform: a fast rsatron is important mainly in light of fast rejection of crapola sent by enemy, rather than for payload per se. [14:59]
* asciilifeform will bbl [15:02]
mircea_popescu: asciilifeform disk afaik fine probably bdb itself shat it at some prior point, tho atm dunno a mechanism. [19:01]
asciilifeform: mircea_popescu: that's the interesting imho part. ftr i have yet to witness a corrupted-but-openable bdb on machine other than where ssd was found to have rotted. [19:01]
asciilifeform: ( there -- found this twice, to date. ) [19:01]
mircea_popescu: instrumentation to hunt for this is by now in the vein of "debug bdb". which im not particularly inclined to do, marked for complete removal anyway. [19:02]
asciilifeform: i'd look at the realloc counter on the disk, betcha it's started to turn. [19:02]
* asciilifeform concurs that there's 0 point in debug of bdb, it's medicine for a corpse [19:03]
mircea_popescu: afaik ssd realloc counter starts turning ~the day they're deployed. [19:03]
asciilifeform: nah [19:03]
asciilifeform: not unless yer using rubbish ssd, at any rate [19:04]
mircea_popescu: in principle, for as long as that's not run out, the disk won't lose sectors. but then again, a definitive answer to this turns into debug-ssd-fw+kernel, which... [19:05]
asciilifeform: (tho, will add, the retarded design of bdb, gives early death on ssd. in sane indexer there is never any reason to touch any single block moar than once ) [19:05]
asciilifeform: in asciilifeform's o(1) tx indexer ( will be welded to an experimental bdb once i get the mmap thing resolved ) there's a 2-level storage -- a 'write-once' o(1) index for blox of age N ( N can be 100-500 in practice ), and a much smaller rewritable one kept strictly in ram ( for 'recent' blox, where the longest chain is potentially movable ) [19:07]
asciilifeform: *experimental trb, err [19:07]
asciilifeform: this gives the 'write block on disk only 1nce' behaviour. [19:08]
asciilifeform: the algo for this, i described coupla yrs ago here. [19:09]
mircea_popescu: http://btcbase.org/log/2019-01-20#1888589 << specialist machine market really died in the 90s, and was buried slowly hence. [19:09]
a111: Logged on 2019-01-20 19:12 bvt: are such machines even build these days? i.e. for dsp, or other use-cases? [19:09]
asciilifeform: mircea_popescu: it's alive, in a sense (various image processing junk) but not sumthing you'd want anyffin to do with i suspect [19:10]
asciilifeform: (e.g. the recent fad for 'face recognizer' even revived 1980s item, neural network processor. i haven't personally tried'em for anyffin , of yet ) [19:11]
mircea_popescu: even the http://trilema.com/2016/give-computers-the-vote-theyre-cheaper-than-women-even/ hollywood production moved to x86. [19:11]
mircea_popescu: asciilifeform only "revived" as a money sucker, afaik deployments still on off the shelf micro [19:11]
asciilifeform: nah see, they want face reader in pnoje, toilet, ballpoint pen, etc. where (despite mighty effort) x86 dun go. [19:12]
mircea_popescu: "want". so far that thing only keeps "researchers" mortgages paid. [19:13]
asciilifeform: ( and again will note, i haven't tried the various irons, may well be snake oils. ) [19:13]
mircea_popescu: w/e, perhaps i do not see teh future. [19:13]
mircea_popescu: ~same utility as quantum computers, as far as i know. [19:13]
asciilifeform: intel for instance sells 1 , for about 20bux in qty. what is actually on the die , i have nfi, and it only worx with their oddball closed compiler thing. [19:13]
asciilifeform: supposedly fits trainable nn, coupla 10k units worth. [19:14]
mircea_popescu: http://btcbase.org/log/2019-01-20#1888626 << you can, because it's so highly structured. [19:14]
a111: Logged on 2019-01-20 19:47 bvt: you can't verify output of 'write 4096 adds' by hand, though [19:14]
mircea_popescu: asciilifeform and "russia" sells "own processor". "only worx" etc. [19:15]
mircea_popescu: http://btcbase.org/log/2019-01-20#1888637 << this is very precisely on point, hence why http://btcbase.org/log/2019-01-20#1888551 is such win and so on. [19:16]
a111: Logged on 2019-01-20 19:59 asciilifeform: a fast rsatron is important mainly in light of fast rejection of crapola sent by enemy, rather than for payload per se. [19:16]
a111: Logged on 2019-01-20 17:20 mircea_popescu: apparently 200 reqs ~per 100ms~ (that's how long it takes to complete one) is not enough to bring it down. [19:16]
asciilifeform: mircea_popescu: y'know, neural net, the thing that classifies pig as 'porn' etc. i'm even willing to believe that the chip worx to spec. not magic by any means, 1970s tech sitting on modern ic [19:16]
mircea_popescu: i know, i know. [19:16]
mircea_popescu: was all the rage when i was in school. alongside the period "GREAT DISCOVERY" of "3d glasses". [19:16]
asciilifeform: i simply can't bring myself to buy any to test, i dun have any pigs that need autoclassifying [19:17]
asciilifeform: ( they offer a 50 $ dev kit, compleete with cd fulla winblowz shitware ) [19:17]
mircea_popescu: (frosi, teh magazine of teh ddr's own pionierorganisation iirc, included a pair in one of the editions i got. i even played with them. [19:18]
mircea_popescu: spoiler : they sucked.) [19:18]
asciilifeform: re nn in general -- i vaguely suspect that a quality rng might actually cure the http://btcbase.org/log/2017-07-20#1687624 problem. but again i dun have any pigs, so haven't had occasion to try. [19:19]
a111: Logged on 2017-07-20 19:52 asciilifeform: because unlearns faster than learns. [19:19]
mircea_popescu: in fact, that publication enduringly shook my confidence in german intelligence. included ALL SORTS of utterly retarded shit -- say a "boat" consisting of a piece of paper with a cut inside, you pour oil there and it "propels" it [19:19]
asciilifeform: mircea_popescu: funnily enuff, decent quality konsoomer '3d visor' appeared , i got one as a gift not long ago. but i dun do chemical models any moar, and so no constructive use for it any moar [19:19]
mircea_popescu: at least the french socialists entirely dedicated themselves to "aesthetics" ie masturbation, rather than "science" and "engineering" [19:20]
mircea_popescu: asciilifeform at the same time, we had very good quality 3d chemistry modeller. it consisted of a bunch of sacks, with colored spheres with holes correctly planarly inclined, and plastic sticks. [19:21]
* asciilifeform as a boy built the classic toy boat from tin can and soldered ballpoint pen tune (the iron kind), add a camp stove pellet and a++ 'pulse jet' [19:21]
mircea_popescu: afaik to this day unbeatable modeller. [19:21]
asciilifeform: mircea_popescu: these are great, but when yer doing a protein gets expensive and bit unwieldy ( picture room-sized stick model ) [19:21]
mircea_popescu: turns out there's NOT THAT MANY possible atomic states. [19:21]
mircea_popescu: asciilifeform no, see, what you call "unwieldy" is the very quality of the item in question. that is it's true value. [19:22]
mircea_popescu: to protest its carnality is as protesting cunt's sliminess. it's "unwieldy" specifically because it works well. [19:22]
asciilifeform: mno. when you have 10,000 sticks, thing sags under gravity, in very 'unphysical' way re the item being modelled. [19:22]
asciilifeform: there's a practical limit to anyffin. [19:22]
mircea_popescu: your head will fall out through your arsehole if you attempt to meaningfully interact with something as large as to make that thing collapse into degenerate matter through a "3d visor" color interferer thingee. [19:23]
mircea_popescu: hm. did that sentence pass parser ? [19:23]
asciilifeform: it did [19:23]
asciilifeform: i used to do this for a living, tho. ( and no it aint a great biz to be in, the models just barely work ) [19:24]
asciilifeform: imho it's a wide green field for hypothetical resurrection of ~analogue comp~ [19:25]
asciilifeform: von neumann comp spectacularly retarded for the app [19:25]
asciilifeform: ( consider, grunting through each atom, serially ) [19:26]
asciilifeform: i had a 30k$ sgi 3d viewer thing, worked ok ( 1 part 50kg crt, other part went on yer head. ) point is that nao you can get ~same thing for 100. not that there's much in the way of people who are still doing any of this afaik. [19:27]
mircea_popescu: i tell you honestly, if i ran a lab where such a compound was seriously important, i'd spend the 100k or w/e it is to have a model welded out of steel balls and rods, and place it in the lab gym. [19:28]
mircea_popescu: let all the bright minds monkey swing from the fucking thing until they fall over. [19:28]
asciilifeform: mircea_popescu: watson & co tried this. [19:29]
asciilifeform: problem is that you want flex in the bonds ( to model strain, van der waals, etc ) but you dun want it to sag. [19:29]
asciilifeform: ( on top of this, the actual meat being modelled is not a rigid object, it has substantially variant geometry depending on temperature and properties of medium, e.g. ph ) [19:31]
asciilifeform: sorta what keeps the old-fashioned crystallography people in biz. [19:31]
asciilifeform: ( btw does mircea_popescu have a theory re why people become crystallographers ? it's almost 100% young chix, and they -- even with modern shielding and safety derpisms -- never live particularly long. multi-MeV xray cannon in yer face... ) [19:32]
asciilifeform: on a few occasions i had to go to where it was done. power supply cable as thick as my thigh. [19:34]
mircea_popescu: i have nfi. [19:34]
mircea_popescu: possibly they tried the dating market. [19:34]
asciilifeform: lol [19:34]
BingoBoingo: Likely [19:35]
asciilifeform: there were even some folx doing protein ~neutronography~. and aaalso chix... [19:36]
asciilifeform: i haven't seen with own eyes tho ( and would rather not ) the biznis-end of that. [19:36]
mircea_popescu: i don't get why they can't simply use a mirror / scintillator / something. [19:37]
asciilifeform: mircea_popescu: they do. [19:37]
asciilifeform: problem is the 'xray hygiene' thing, outlined by mircea_popescu in earlier thread [19:37]
mircea_popescu: myeah... [19:37]
asciilifeform: http://btcbase.org/log/2018-09-11#1849389 << oblig slotin oldthread. [19:43]
a111: Logged on 2018-09-11 01:22 asciilifeform: incidentally much noise was made re the elaborate calculations of early nukes, but good chunk of the design params ~also~ only were accessible empirically. hence the slotin incident. [19:43]
mircea_popescu: while the former is actually true, the latter doesn't follow, it's not like he had to use the beryl for a soup bowl. [19:44]
asciilifeform: well arrhenius or who it was, didn't ~have~ to taste his reaction products either [19:44]
asciilifeform: but some folx, when considering 'hmm, taste, or go home' will taste. [19:45]
mircea_popescu: you telling me slotin was a sort of curie jr, had "no idea what could occur" ? [19:45]
asciilifeform: nah [19:45]
asciilifeform: extreme case imho. but illustrative of principle. [19:45]
mircea_popescu: well then. monkey was monkying about, gimme a break. [19:45]
mircea_popescu: 0 respect for any kind of safety. [19:45]
asciilifeform: eh life was cheap, electricians routinely measured voltage with wet finger etc. [19:46]
asciilifeform: even as late as '90s lab folx had to be beaten with a stick to get'em to stop mouth-pipetting, etc [19:46]
BingoBoingo: Putting on a lab coat doesn't make a monkey hygenic [19:47]
asciilifeform: ( i suspect that current-day 'millenial' derps would shit their pants if they saw folx actually at work and the things that are done even nao ) [19:47]
mircea_popescu: yeah... you know i used to do the old thumb-stopper thing... until i noticed i got spots. [19:48]
mircea_popescu: BUT I WAS THRITEEN [19:48]
asciilifeform: mircea_popescu: i was cured of this by chem instructor, who showed palm where glass pipe went clean through it when he was a lad [19:48]
asciilifeform: little round scar on ea. side [19:48]
mircea_popescu: heh. [19:49]
asciilifeform: i dun consider slotin et al to be shining examples to follow, but will also point out , that errybody dies eventually, and personally would rather chance to eat 100kV than to die of boredom sitting in a chair kanzure-style. [19:52]
* asciilifeform was thinking of this just this morning, while varnishing the 50kv cabling on the xray box [19:53]
mircea_popescu: by the time your evening's entertainment is screwdriver-juggling neutron reflectors, you've got some pretty serious pretentious issues going. i mean aestheticism is one thing, but by the time that's all that's left... [19:53]
asciilifeform: as i understand, usg dept. of wunderwaffen hasn't given'em robo-cranes yet, at that pt [19:54]
mircea_popescu: there's space in the middle. [19:54]
asciilifeform: surely. [19:54]
mircea_popescu: meanwhile on the other side, "About me I love the X Files and Criminal Minds. I'm studying Cosmetology right now but next semester I am transferring to Psychology. " [19:57]
asciilifeform: now see ~some~ people, odin himself made'em to prop up neutron reflector with dainty little fingers. [19:58]
asciilifeform: they are simply misplaced atm. [19:58]
mircea_popescu: sadly, their way to their butugychag is blocked by a bunch of shriveled old prunes decked in star trek paraphenalia. [19:58]
asciilifeform: tragic waste. [19:58]
asciilifeform: sadly. [19:59]
mircea_popescu: i suppose you might enjoy the whole thing. http://p.bvulpes.com/pastes/hYrat/?raw=true [19:59]
asciilifeform: 'However I’m very into Financial Domination and paying me is the only way' << lol!! [19:59]
mircea_popescu: somebody asked her, see. [20:00]
asciilifeform: i can't resist to ask, is this one sui generis or is 'financial domination' actual button on the fatlife dashboard thing nao. [20:01]
asciilifeform: re 'somebody asked', i suspect that the correct physical model is that old toy ( do they still make'em ? ) with plastic phonograph roll inside, where you pull string and it 'speaks' [20:02]
asciilifeform: you pulled, see, its string. [20:02]
asciilifeform: so it 'spoke'. [20:02]
BingoBoingo: asciilifeform: It's an actual thing in the sense "black world order" is a thing [20:03]
asciilifeform: i gotta take off hat to mircea_popescu's strong stomach, i can't picture fucking these things even with ball gag. would almost rather juggle neutron reflectors. [20:04]
mircea_popescu: sui generis. there's a whorde of these cosmetologists with aspirations, populating what's left of the human side of usg's "ugc" lines, filling up the "journalist" pipelines, camwhoring etc. [20:04]
mircea_popescu: asciilifeform this particular one entirely unfuckable, tiny tits an' a belly. but in general, idiots have ~no chance of making it through anyway, so can't say as i have. [20:05]
BingoBoingo: "I like being treated like a princess and fucked like a slut." << The anglophone girl problem. They think they want to hold the power, but they know they want the power imposed on them when it counts. [20:06]
mircea_popescu: i do recall in youth was a lot less selective/priviledged, whichever way you look at it but somehow inexplicably stupidity not nearly as organised or militant in same youth. i mean, fuck, girls back then were ASHAMED to sit with the mainstream, jesus fuck. [20:07]
asciilifeform: BingoBoingo: i'm not even convinced there's an 'i' in that there crankcase that is capable of actually 'wanting' sumthing. [20:07]
mircea_popescu: so i guess i switched historical priviledge for social priviledge as i aged. [20:07]
BingoBoingo: asciilifeform: Well, two decades of child abuse organized for her by Disney [20:08]
mircea_popescu: can't say that there's what to dispute this view with, that they've failed to meaningfully individuate. [20:09]
asciilifeform: aha. and there aint a cure, any moar than for hydrocephalia. the ideal life outcome for these, is 'institutional' , the kind with shock collar, detonguing, short leash, i suspect. [20:10]
mircea_popescu: well, if hillary had her way they'll all be "social "workers" coming up with rules an' regulations for each other. [20:11]
mircea_popescu: as it is... i dunno, i suspect they'll starve, orange man doesn't seem inclined to continue the slop. [20:11]
mircea_popescu: but for instance i have standing orders for 0 tolerance and active harassment / belittling and sabotage of them by my girls, whenever encountered (which mostly is receptionists, but hey) [20:12]
BingoBoingo: If UruguArgenZuela is illustrative they'll fad their way into avoiding starvation with "coliving" and food "recollection" [20:13]
mircea_popescu: rapidly becoming the social pariah, this cosmetologist princess-psychologist. [20:13]
asciilifeform: i had nfi the species is found as far south as mircea_popescustan [20:13]
mircea_popescu: principally because they're so easy to filter them out. [20:13]
mircea_popescu: asciilifeform are you kidding me ? the species is found anywhere samsung put "smart"phones on layaway. [20:14]
mircea_popescu: like plague follows rats, these follow touchscreens [20:14]
asciilifeform: hm i thought they could only grow where a similarly-porcine mother lives as 'office plankton' and prevents contact w/ reality [20:15]
mircea_popescu: the "billions" that schmuck was talking about. [20:15]
BingoBoingo: asciilifeform: at last 1/3 of the pichis here have pnoHes [20:15]
mircea_popescu: asciilifeform nah, generally parents content to "not understand" offspring and "feed, anyway" [20:15]
mircea_popescu: this is pretty fucking amusing to me, too. "oh, i don't understand the kids... and this is my problem". bitch... if i don't understand them i don't feed them, this "you don't understand me" is just about "tomorrow, i die", naught else. let them fucking work ass off to be understood, wtf, you want me to do it ?! [20:16]
mircea_popescu: vaguely reminescent of occasional fetard going "i don't know you" / occasional moron going "this trilema article is too long and complicated for my understander". [20:16]
mircea_popescu: "what the fuck do you expect ME to do about that ?!" [20:17]
mircea_popescu: the expectation that a breakdown in communication is a stable state better come with a baked-in expectation of 0 financial flows, wtf. [20:18]
feedbot: http://pizarroisp.net/2019/01/21/pizarro-isp-update-january-21-2019/ << PizarroISP -- Pizarro ISP Update January 21, 2019 [20:20]
asciilifeform: BingoBoingo: battery consistently in mid-60s ( all-time high 68 ), cpu - 50s [20:24]
mircea_popescu: could a very shitty battery meltdown and fuck the rk then ? [20:25]
asciilifeform: mircea_popescu: the rk case is about 1.5 metre away [20:25]
asciilifeform: BingoBoingo: i'ma include a simple machine-readable ambient thermometer/hygrometer in the cargo, for good measure [20:26]
mircea_popescu: oh, so this is the shared ? [20:26]
asciilifeform: mircea_popescu: the temps/voltages thrd is re 'uy1' , the collective www box yea [20:27]
mircea_popescu: aha [20:27]
asciilifeform: it's a 1u opteron box with 32 cpu [20:27]
asciilifeform: ( of same type as dulap-iii and the s.mg machines ) [20:28]
asciilifeform: runs off 2 physically separate 220v ps. [20:29]
asciilifeform: BingoBoingo: btw see if you can hunt down a mains quality recorder that doesn't cost own weight in gold, so far my dig has disappointed [20:31]
asciilifeform: ( hunt on net, that is, not in BingoBoingostani junkyard , tho i'll be pleasantly surprised if it were to turn up ~there~ ) [20:32]
* asciilifeform will bbl : meatspace subsystems [20:33]
BingoBoingo: asciilifeform: I'll add it to the ask list when I make the rounds to the shops [20:50]
BingoBoingo: In addition to net searching [20:54]
mircea_popescu: !!up fromdeedbot [21:49]
deedbot: fromdeedbot voiced for 30 minutes. [21:49]
mircea_popescu: meanwhile in historical problems, http://btcbase.org/log/2014-08-30#816689 [23:16]
a111: Logged on 2014-08-30 23:24 mircea_popescu: i have a glass table and an optic mouse. [23:16]
asciilifeform: mircea_popescu: interestingly, recent ones, with ir laser, actually work on glass.. [23:29]
mircea_popescu: i thought lasers didn't exist! [23:29]
asciilifeform: lol [23:29]
mircea_popescu: :p [23:30]
mircea_popescu: anyway, i was looking for the story of the mouse chewing at the jar lid hole with his tiny teeth [23:30]
mircea_popescu: but can't fucking find it anymore. [23:30]
mircea_popescu: fat people can't find their feet but i'm so overweight i can't find my fucking stories. [23:31]
asciilifeform: ( exercise for reader : what is won by using laser diode in mice, vs ordinary led ? ) [23:31]
asciilifeform: !#s mouse jar [23:31]
a111: 13 results for "mouse jar", http://btcbase.org/log-search?q=mouse%20jar [23:31]
mircea_popescu: not there sadly. [23:32]
asciilifeform: http://btcbase.org/log/2014-09-07#821631 [23:35]
a111: Logged on 2014-09-07 17:02 mircea_popescu: something like that. so one night while in here, i see movement in the corner of my eye [23:35]
mircea_popescu: aaaa! TY! [23:36]
mircea_popescu: that's the fuck it was, 2014. [23:36]
BingoBoingo: So, total lunar eclipse balcony view. Pretty cool, but lacks the magic of solar totality. [23:48]
mircea_popescu: harem vote decided we're too lazy to go check it out. [23:49]
BingoBoingo: Well, I used a compass and a map to determine I didn't have to go anywhere special to see it and thus it was seen [23:51]
mircea_popescu: pretty sure it's partial here, and i'm on the wrong side of teh hill momentarily. [23:51]
BingoBoingo: I'd go out of my way to see a total solar eclipse again, very strong inducement of wonder. Lunar well... not so much [23:55]
BingoBoingo: And the moon going into and out of totality isn't the shocker that the sun is. At least not when artificial lighting is involved. [23:57]
Category: Logs
Comments feed : RSS 2.0. Leave your own comment below, or send a trackback.
Add your cents! »
    If this is your first comment, it will wait to be approved. This usually takes a few hours. Subsequent comments are not delayed.