log☇︎
42300+ entries in 0.022s
asciilifeform: i'ma detail, ftr : 'ffacalc' runs 'as fed', i.e. 1 command at a time. but 'peh' , adult version, has support for functions and loops, and therefore requires the 'tape' to exist in memory. so currently i have 'tape can be 1000000 bytes', but this is not acceptable obv. in the long term
asciilifeform: diana_coman: the 'errything on stack' approach has its limits; it is why i wrote the mmap thing (currently stuck in limbo , but i'ma have to revive it and fix, cuz ffa 17 also is hitting against this wall, you can't expect to put 100MB on stack, you gotta mmap it
asciilifeform: diana_coman: i think he's trapped in some sorta cube hell; the squarebracket thing mircea_popescu asked for also not happened yet etc
lobbesbot: asciilifeform: The operation succeeded.
a111: Logged on 2019-03-01 08:57 phf: asciilifeform: give me until end of march to resolve it one way or another, feel free to neg rate me then
asciilifeform: !Q later tell phf fughet for nao about http://btcbase.org/log/2019-03-01#1899897 , but wouldja pleez fix vdiff ? and tell us what sorta swamp yer stuck in, what wouldja need to get to surface ? ☝︎
diana_coman: (i.e. it runs, it returns fine, no differences between the files)
asciilifeform: so hard to say 'on napkin' what mass chokes it
diana_coman: mircea_popescu, this is a different machine, the cuntoo-guinea pig
asciilifeform: diana_coman: iirc it throws whole file on the stack, but then also keccak eats stack
mircea_popescu: diana_coman try it with the megastack size from before, calling tests ?
mircea_popescu: asciilifeform that was size, this is ???
asciilifeform: i'd like to see phf come back to life and fix. failing this, 1 of us will have to ☟︎
asciilifeform: diana_coman: i do not presently know where is the barf threshold, i suspect it depends on yer stack size
diana_coman: and yes, I had this idea in my head that "previous problem, was solved"
diana_coman: asciilifeform, ftr this is a 4.7MB file
asciilifeform: phf promised to fix, but then went on his ill-fated voyage
a111: Logged on 2018-10-20 01:44 asciilifeform: !Q later tell phf i found today that your keccak-vdiff is unable to eat a 40MB file ( dies politely with stack overflow )
asciilifeform: mircea_popescu: we found that vdiff overflows during http://btcbase.org/log/2018-10-20#1864346 ☝︎
asciilifeform: ( the style of programming that would appear on ' asciilifeform's ideal cpu ' is best illustrated in http://btcbase.org/patches/fg-genesis/tree/fg.v . i.e. all of the independent pieces in fact run in parallel, and in deterministic time, there are no interrupts, no scheduler, etc. )
diana_coman: mircea_popescu, yes, vdiff on the 2 genesis.vpatch files overflows the stack; while on same machine and same files, diff seems to have no such trouble
asciilifeform: http://btcbase.org/log/2019-03-10#1901140 << the down side of 'let's 512b bus' is that most cpu time (where it runs, not counting idles on i/o here) is spent in 'inner loops' where yer counting to e.g. 3. and nao you gotta move 512bits when yer counting to... 3 ☝︎
a111: Logged on 2019-03-10 17:02 mircea_popescu: perhaps the correct republican approach is not to bake cpu, but to ~bake memory~. why even bother with the whole turdpile that's ddr init when could simply have sane ram, and rk with it.
asciilifeform: http://btcbase.org/log/2019-03-10#1901153 << bake 'pile of reconnectable flipflop' and then you aint gotta ever bake anyffin else again. iirc i detailed this in ancient thread, mircea_popescu barfed ( iirc answered 'why waste so many transistor on interconnects' ) , but can't currently dig up where we had this ☝︎
a111: Logged on 2019-03-10 16:43 mircea_popescu: ie, that structured data should really be much better hardware supported, that memory should include much more processor per storage cell, that in fact memory should look a lot more like trees than the current flat, democratic lines of "all cells are equal"
asciilifeform: http://btcbase.org/log/2019-03-10#1901148 << imho the ( ~homogeneous~ variant of ) fpga is actually the correct model. i.e. you get to stitch it later into however many parallel mechanisms you happen to need on a given occasion. ☝︎
a111: Logged on 2019-03-10 16:42 mircea_popescu: asciilifeform thinking about whart you said, i can't repress the suspicion that maybe the memory model is acrtually profoundly fucked,as a central driver of the whole cs insanity.
asciilifeform: http://btcbase.org/log/2019-03-10#1901147 << it indeed is, and in precisely the way described. ☝︎
a111: Logged on 2019-03-10 08:23 mircea_popescu: asciilifeform so i take it your ideal cpu would actually be simply state machines + registers, no actual ram ?
asciilifeform: http://btcbase.org/log/2019-03-10#1901142 << programmable interconnect fabric ( similar to what's sold as fpga ) . iirc i detailed this in old thrd. ☝︎
mircea_popescu: make a 18446744073709551616 byte ram arm board, for the keks.
a111: Logged on 2018-09-25 00:39 asciilifeform: the only binball is that coupla kB of ddr ram init thing.
mircea_popescu: perhaps the correct republican approach is not to bake cpu, but to ~bake memory~. why even bother with the whole turdpile that's ddr init when could simply have sane ram, and rk with it. ☝︎☟︎
a111: Logged on 2019-03-10 15:10 diana_coman: onth in unexpected results and assorted ugh: vdiff ends up in stack overflow ran on those
mircea_popescu: http://btcbase.org/log/2019-03-10#1901145 << say this again ?! vdiff blows the stack when processing the two diffs ? ☝︎
mircea_popescu: ie, socialism fails yet again, and calling itself "democracy" dun help anything -- flat is fail, no two ways about it.
mircea_popescu: ie, that structured data should really be much better hardware supported, that memory should include much more processor per storage cell, that in fact memory should look a lot more like trees than the current flat, democratic lines of "all cells are equal" ☟︎
mircea_popescu: asciilifeform thinking about whart you said, i can't repress the suspicion that maybe the memory model is acrtually profoundly fucked,as a central driver of the whole cs insanity. ☟︎
diana_coman: trinque, which are exactly the paths that don't match? since I don't have the original genesis.vpatch I can't really know what to check to look if indeed those paths actually exists or not or wtf
diana_coman: onth in unexpected results and assorted ugh: vdiff ends up in stack overflow ran on those ☟︎
a111: Logged on 2019-03-09 22:43 trinque: diana_coman, other folks that have cuntooed, can y'all confirm that the paths that ended up in your genesis.vpatch do not in fact exist? I'd like you to reproduce the commands starting at line 114 of scripts/make_portage_tree.sh in your build directories, i.e. cd ~/src/cuntoo/build/cuntoo and then run them, as root
diana_coman: http://btcbase.org/log/2019-03-09#1901066 -> I ran those and I got exactly the same genesis.vpatch (i.e. diff on this vs the one obtained from the script itself returned empty) ☝︎
mircea_popescu: asciilifeform so i take it your ideal cpu would actually be simply state machines + registers, no actual ram ? ☟︎
a111: Logged on 2019-03-10 03:21 asciilifeform: http://btcbase.org/log/2019-03-10#1901103 << i cannot resist to bite : why 512 ?
mircea_popescu: o right the registers. i recall.
asciilifeform bbl, maffs then meat
asciilifeform: i dun expect to even live to see with own eyes a machine where 64bits of addr space fully populated (on current x64 Official standard, only 48 addr lines even connected, the rest mandatory 0)
a111: Logged on 2019-03-10 03:02 mircea_popescu: memory, of course, starts with byte 299008 (584 * 512) and extends as far as 512 bits bus can address, if you're installing that. meaning, all core registers are allocated as memory anyways. and that's fucking it.
asciilifeform: http://btcbase.org/log/2019-03-10#1901107 << 2^512 bits weigh approx what galaxy weighs, try it ☝︎
a111: Logged on 2019-03-10 02:23 mircea_popescu: vp of motherfucking what, exactly ? "how to give away the market to the azns" ? that's a skill ?
asciilifeform: http://btcbase.org/log/2019-03-10#1901094 << with or without crapple, asia laffs allthewaytothebank -- it aint as if crapple bricks are made somewhere other than shenzhen ☝︎
a111: Logged on 2019-03-10 02:24 mircea_popescu: any fucking monkey picked out in the street could have made ~just as good~ a "corporate executive" as the fucktards apple hired -- and somehow nobody is telling them this. and of course the idiots don't have the werewithal to look in the mirror, "if plumbers were as good at plumbing as we are at executiving, we'd be quite literally swimming in the shit we're metaphorically drowning in!"
asciilifeform: http://btcbase.org/log/2019-03-10#1901095 << rando monkey from the zoo would make 9000x better director than the catamite d00d ( recall, what he was in charge of prev. : rounding the corners on the 'ui'.. ) ☝︎
asciilifeform: aaaanyfffing but get rid of the idjit von neumann bottleneck and bake micro-cpus into the ram
asciilifeform: so they 'let's find sumthing to do'.
a111: Logged on 2019-03-10 03:04 mircea_popescu: i ~can't imagine~ what the fuck must have been going through the skull of whoever came up with "working a piece or moving the worked piece, same thing". what the fuck ever is it same thing!
asciilifeform: http://btcbase.org/log/2019-03-10#1901108 << pipelineism, branchpredictionism, etc., all these heresies, were birthed from the fact that speed of memory fell massively behind that of cpu, in the time it takes to fetch ~anything~ you can do five digits of clock cycle ☝︎
asciilifeform: the thing with gigantic multers is that they grow physically with the cube of the bitness. hence scarce. ( tho i dun imagine even a 8192bit single-cycle multer would be remotely near as heavy as the 3bil-transistor 'let's fry eggs' pentium-xxviii or whatnot )
a111: Logged on 2019-03-10 03:02 mircea_popescu: the 64 "flat outer" cores are state machines that can do mvin and mvout -- taking registers to memory or memory to registers, ie use the bus. only, of course, for the state machines inside their ~projection~.
asciilifeform: http://btcbase.org/log/2019-03-10#1901105 << this is sorta how hillis's 'connection machine' worked. was pretty neat. ☝︎
a111: Logged on 2019-03-10 03:02 mircea_popescu: but speaking of those design decisions, by my current thinking the ideal processor is defined as follows : the bus width is 512bits ; therefore the byte is 512bits. the processing core is a state machine, with 512 byte-sized registers. a processor is composed of 512 + 8*9 such cores. for convenience imagine them organized in a cube, 8x8x8, with an extra 64 item layer on three sides.
asciilifeform: http://btcbase.org/log/2019-03-10#1901103 << i cannot resist to bite : why 512 ? ☝︎☟︎
asciilifeform: he's on ch9, where no barrett, ~80% of the cpu eaten by knuthian div.
a111: Logged on 2019-03-10 02:46 mircea_popescu: 7% gains, not even that huge.
asciilifeform: http://btcbase.org/log/2019-03-10#1901102 << see the comments ☝︎
asciilifeform: think, where else are idjits gonna ~lease~ ( the new crapple thing! not even buy, lease.. ) a brick for its weight in gold
asciilifeform: ( possibly am mistaken about this ? )
a111: Logged on 2019-03-10 02:22 mircea_popescu: http://btcbase.org/log/2019-03-09#1901064 << in quite related news : went to flagship mall in this country, escazu multiplaza (escazu being where the us embassy compound lies, and all the retarded gringos live, very miami real estate racket reservation), and guess what ? there's no ipad store. AT ALL. instead, huawei dominates both in advertising (these large floating banners) and location (top floor store, only one to sell c
asciilifeform: http://btcbase.org/log/2019-03-10#1901090 << i was convinced somehow that crapple stores exist only in the reich ☝︎
a111: Logged on 2019-03-10 02:44 mircea_popescu: http://bvt-trace.net/2019/03/ffa-chapter-9-homework-comba-in-x86_64-assembly/#selection-15.295-19.176 << why this, specifically ? is there no ada asm calling method besides this ?
asciilifeform: http://btcbase.org/log/2019-03-10#1901101 << recall, ave1 found that the asm inlining was ~yet another~ item partially broken in ye olde gnat ( iirc it dun let you assign registers deterministically ) ☝︎
mircea_popescu: i ~can't imagine~ what the fuck must have been going through the skull of whoever came up with "working a piece or moving the worked piece, same thing". what the fuck ever is it same thing! ☟︎
mircea_popescu: memory, of course, starts with byte 299008 (584 * 512) and extends as far as 512 bits bus can address, if you're installing that. meaning, all core registers are allocated as memory anyways. and that's fucking it. ☟︎
mircea_popescu: the remainder 8 "sharp outer" cores control the flats, by moving things in and out of ~their~ registers.
mircea_popescu: the 64 "flat outer" cores are state machines that can do mvin and mvout -- taking registers to memory or memory to registers, ie use the bus. only, of course, for the state machines inside their ~projection~. ☟︎
mircea_popescu: the 512 "central" cores are state machines that can do add or mul, and always proceed ~on their entire register set~. so if you don't want to multiply 131072-bit numbers, just put in zeros ; and if you put larger numbers in there you'll just get the LAST 262144 bits of the result, is all.
mircea_popescu: but speaking of those design decisions, by my current thinking the ideal processor is defined as follows : the bus width is 512bits ; therefore the byte is 512bits. the processing core is a state machine, with 512 byte-sized registers. a processor is composed of 512 + 8*9 such cores. for convenience imagine them organized in a cube, 8x8x8, with an extra 64 item layer on three sides. ☟︎
mircea_popescu: 7% gains, not even that huge. ☟︎
mircea_popescu: http://bvt-trace.net/2019/03/ffa-chapter-9-homework-comba-in-x86_64-assembly/#selection-15.295-19.176 << why this, specifically ? is there no ada asm calling method besides this ? ☟︎☟︎
mircea_popescu: but on a more optimistic note, that's probably the most anyone cared about anything to do with the topic or the persons named or otherwise referred to in their entire history. that should count for something.
mircea_popescu: dude check me out, by now i'm writing 500 word chatlines. this isn't going well.
mircea_popescu: id cow "from twitter" explaining @whatever conference "how serious they take banning" blinks incomprehendingly just as she's done with her soundbytes. yeah, why is it ??? there's entirely no difference between any of them and any other aspiring-writers-in-new-york, philosophy of art history studies rejects the world over. why those, why not these ? blink blink ?
a111: Logged on 2017-03-15 00:29 mircea_popescu: pretty fine example of exactly why warren was so vocal (item was strictly a barony created so elizabeth warren could be barron OF SOMETHING). this cfpb item spent 55mn on "renovations" of its hq, ie more than the gsa spent that year on everything the usg owns ; spent immensely on travel (which is not something they do). the chairman is supposed to not be removable by the president except "for cause" (meanwhile that got strick
mircea_popescu: this is the fucking problem of socialism, when that wanna-be alt-hilary stupid cow asks "if the government can print money to rescue wall street, why won't it print money to let the chitlins enjoy the college lifestyle free of charge (and permanently!)" she has a fucking point -- and the entirely similar stup ☝︎
mircea_popescu: any fucking monkey picked out in the street could have made ~just as good~ a "corporate executive" as the fucktards apple hired -- and somehow nobody is telling them this. and of course the idiots don't have the werewithal to look in the mirror, "if plumbers were as good at plumbing as we are at executiving, we'd be quite literally swimming in the shit we're metaphorically drowning in!" ☟︎
mircea_popescu: vp of motherfucking what, exactly ? "how to give away the market to the azns" ? that's a skill ? ☟︎
mircea_popescu: jobs' been dead what, a decade ? not even a decade. meanwhile, fifty FUCKING MORONS sat around in rooms pompously pretending as to how "of course i'm teh vp, didn';t you see the sign on the door ?"
a111: Logged on 2019-03-09 22:40 asciilifeform: when you build 1 of these things, there's a set of decisions that end up determining shape of whole thing; and it so happens that intel made ~all~ of the most retarded possible choices.
mircea_popescu: http://btcbase.org/log/2019-03-09#1901064 << in quite related news : went to flagship mall in this country, escazu multiplaza (escazu being where the us embassy compound lies, and all the retarded gringos live, very miami real estate racket reservation), and guess what ? there's no ipad store. AT ALL. instead, huawei dominates both in advertising (these large floating banners) and location (top floor store, only one to sell c ☝︎☟︎
mircea_popescu: 000000000000101!!! NOW THAT IS IMPORTANT!!!
a111: Logged on 2019-03-09 22:38 asciilifeform: if you ever wonder why your x64 iron draws 50x the wattage to do same thing as e.g. rk, wonder no longer -- the insanity where shit gets moved around to accomodate idjit instructions with fixed in/out hoppers, the insanity where you gotta set prefixes to specify what width ~each operand~ is (why this is needed ? srsly) , all of this adds up to 3bil transistors that heat the room
mircea_popescu: http://btcbase.org/log/2019-03-09#1901062 << very much this. "oh, couldn't POSSIBLY spare another bus width for a FULL mult result. NEVERTHELESS... can fucking spare eight trillion bus widths to specify instruction length. it's 00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000101 and not a mere fucking 0000000000000000000000000000000000000000000000000 ☝︎
asciilifeform: remember, only a terrorist(tm)(r) 'writes own crypto', 'good citizens use openssl' etc etc
asciilifeform: but no prizes for guessing why it aint on the market.
asciilifeform: ftr an 'iron ffa' cpu does not even require a massive multiplier . even a microcoded ffa-style thing that lets you specify 'and at memory x there is a w-word int, and at y a w word int, add'em' etc , would still massively win over the extant liquishit, it would do the arithm atomically, without invoking branchpredictor, losing cache, etc.
asciilifeform: ... so 'mulx' aint in anyffing i have. if someone wants to test it with own hands, he can, otherwise fughetit.
asciilifeform: and meanwhile2, we have answr to above quandary, 'In 2017, BMI2 was further incorporated in AMD's Zen-architecture...'
asciilifeform meanwhile found today mistake in 17 , and expects it'll take several days to rewrite
asciilifeform: ftr asciilifeform suspects that 99% of what can be won from asmism in ffa, can be had simply from bvt's existing 64bit mul, plus doing adc for the additions-with-carry instead of the manually-cranked carry calc, and that all 'fancy' instructions will only lead to sad
asciilifeform: fwiw it's still a 64bit mul, the only win is that it dun set any flags (and therefore keeps the pipe flowing)