asciilifeform: mircea_popescu: i did conjecture that you can skip normalization if you widen the register.
asciilifeform: given that currently we have only symmetric (equal-bitness multiplicands) multiplication, which will waste massive cycles
asciilifeform: as for gains, i dun know for sure yet, because i gotta manually implement the asymmetric karatsubaization for the above
asciilifeform: he proposed to do the normalization as usual, but with 'secret shift' algo. which , any way you cut it, eats more cycles than knuthian division even.
asciilifeform: mircea_popescu: this is quite different from the apeloyee d00d's suggested method
asciilifeform: in ordinary barretting, x loses its bottom however-many bits, to make the multiplication smaller, depending on the 'bitlength' of x; and then its gets restored with a leftshift . we can't do this, because ffa.
asciilifeform: now r is == to the remainder of x/m .
asciilifeform: so now we gotta r := x - q*m , from 1 to 4 times ( this is done via subtract and mux-on-carry , so it always happens 4 times, but when q becomes == to q' the mux stops changing the result )☟︎
asciilifeform: if q' is the actual quotient, the inequality q' <= q <= 4q' holds. ( proof left for the reader )
asciilifeform: compute the quotient-guess q, q := (c * x) >> 4096☟︎
asciilifeform: however it would very conspicuously rock , imho, if one of you folx did this, instead of waiting.☟︎
asciilifeform: it also occurs to asciilifeform that he is doomed to at some point write something like a sane 'tinyscheme' in ada. so that we have a scripting lang and can finally pour the cement into grave of python.
asciilifeform: afaik 100% of the useful functionality, reduces to 'half-written, buggy implementation of' v.
asciilifeform: trinque: to this day i have nfi why portage weighs what it weighs, or what most of the liquishit does
asciilifeform: 'The US is also in the process of establishing a new drone base just outside the city of Agadez in Niger in an effort to bolster regional counterterrorism efforts.' << mno attacked FORNOREASON!111
asciilifeform: people laughed when they opened 1982 ibm pc and found working dram chips made of 3, 4 semidead ones soldered 'dead bug' on one another. but those WORKED.
asciilifeform: and this is considered salable hardware ?!
asciilifeform: here's a choice lulzgem: 'One-location hammering is based on a previously unknown Rowhammer effect. With one-location hammering, the attacker only runs a Flush+Reload loop on a single memory address at the maximum frequency. This virtually keeps the DRAM bank permanently open. We observed that one-location hammering drains enough charge from the DRAM cells to induce bit flips.'☟︎
asciilifeform: incidentally if you dun let randos twiddler yer address lines ( i.e. run ffaically ) you dun get hammered.
asciilifeform: ( btw is it clear that buildroot is only part of the trb universe because we dun have a musltronic linux to do ordinary stator builds in ? )
asciilifeform: ' 'Why do the nutrition facts tell me there are 25 crackers in a serving, 11 servings to a box? Just give me the nutrition facts for the whole damn box, that’s my serving size.' ' << lol!!
asciilifeform: diana_coman: which part needs bzip ??
asciilifeform: ACHTUNG panzers! does anybody in l1 want a sage probe ( http://www.loper-os.org/?p=1887 ) for anything, and dun have one yet ? i bought one from the d00d in the comments to my orig. article, for next-to-nothing . if any of you lot want it, it's yours for cost of postage
asciilifeform: nigerian prince spam moar interesting.
asciilifeform: ( does mircea_popescu much hang out with folx who see mouse as man ? it isn't a thing that'd normally occur to asciilifeform , say . where does he get the vegetarians )
asciilifeform: q is whether elliot is mouse, not whether mouse is a man
asciilifeform: mouse did not, i suspect, devote rest of his days to smashing glass wherever found.
asciilifeform: sorta like mircea_popescu's jar mouse , only superficially 'hated the glass'
