asciilifeform: aha worx

asciilifeform: mircea_popescu: 2nd eggog, 'very significant costs'

asciilifeform: mircea_popescu: 'underlying problem' link in your latest piece, eggogs

asciilifeform: ^ e.g.

asciilifeform: !A ??``*[print 0x]##[ == 0x]#[ * 0x]#

asciilifeform: there you go.

asciilifeform: aaha.

asciilifeform: !A ??*Q

asciilifeform: !!up pehbot

asciilifeform: and that the python example uses both of'em, to test the entire width of the multiplier. and works.

asciilifeform: observe that * produces ~two~ FZ onto the stack, not one.

asciilifeform: phf: why didja truncate to the ffa bitness ?

asciilifeform: phf: chew this apart for me ?

asciilifeform: ( it gives answer to the 'make auto tests' homework from ch8 )

asciilifeform: mod6: you'll like this one.

asciilifeform: Achtung, Panzerz ! ^

asciilifeform: !~later tell phf plox to snarf ch9

asciilifeform: that's approx it.

asciilifeform: lol

asciilifeform: dunwork so much with parcels in the post. because otherwise 'erryone would write it'

asciilifeform: phf: in my experience the magic word is 'commercial sample for demo'. but only worx with suitcases.

asciilifeform: mod6: at no point does babysitting-and-restarting-blindly accomplish anything useful.

asciilifeform: mod6: at the risk of repeating myself : don't babysit nodes. if it genuinely 'gets stuck', and not merely on account of poor block propagation at the particular time you are watching -- it is potentially serious problem, and you want debug info . if on other hand it is a chronic situation, you want to instrument the thing with timers, find out where it spends time.

asciilifeform: phf: worx with variable success. didn't , for instance, for FG to diana_coman

asciilifeform: from 2005 or whenever.

asciilifeform: ben_vulpes: if you're especially unlucky -- by googling and finding a usg purchase quote from $bigvendor

asciilifeform: rather than constantly

asciilifeform: observe also that even 'aggressive' trb only demands newblox when a peer ~connects~

asciilifeform: gotta be specific re what node is doing. word 'stuck' is not very useful.

asciilifeform: ... or still verifying something, and doing nothing at all, not responding to commands, etc ?

asciilifeform: e.g. sitting and processing tx, and simply nobody gave it any new blocks ?

asciilifeform: and what is it doing

asciilifeform: how long is long

asciilifeform: and fwiw i NEVER restart a node unless it actually crashed ( and it's been some yrs since last case of this ), why would you destroy valuable info re an actual-stuck eggog

asciilifeform: mod6: what means 'stuck on a block'

asciilifeform: naa i'ma do what mircea_popescu said, leaving it alone for nao

asciilifeform: *if i turn

asciilifeform: what would the contrary even mean ? that the low bit is on left side of the chip , vs right ? what if it turn my comp around on the table , lol

asciilifeform: 'bit order' does not meaningfully exist outside of serial lines, where a time parameter is imposed

asciilifeform: well yes. but you gotta use the подвиг радиста UI to get'em..

asciilifeform: http://btcbase.org/log/2014-04-25#643502 << ☝︎

asciilifeform: unless you are that d00d in the painting

asciilifeform: there does not exist this separation, in reality, between display and ontology.

asciilifeform: sad phakt : this is also true on little-endian box -- at least if you watch the bits go over e.g. serial port.

asciilifeform: but we dun have a cpu fab yet.

asciilifeform: i also agree with mircea_popescu in re ~new~ cpu design having no business being byte-addressing and having a detectable endianism orientation AT ALL.

asciilifeform finds that he agrees with phf : like it or not, a good % of the cheap and effective older iron that is and will remain in use in tmsr , specifically in opposition to x86ism, is big-endian.

asciilifeform: diff -uNr $1 $2 | awk 'm = /^(--- a|\+\+\+)/{cmd="sha512sum \"" $2 "\" 2>/dev/null ";s=cmd| getline x; if (s) { split(x, a, " "); o = a[1]; } else {o = "false";} close(cmd); print $1 " " $2 " " o} !m { print $0 }'

asciilifeform: #!/bin/sh

asciilifeform: alternatively,

asciilifeform: not with any of the currently published vdiffs

asciilifeform: and worse, i can't even readily think of a simple pill against this inbandism.

asciilifeform: +-- (C) 2018 Stanislav Datskovskiy ( www.loper-os.org ) --

asciilifeform: --- (C) false

asciilifeform: but this becomes , in vdiff,

asciilifeform: +-- (C) 2018 Stanislav Datskovskiy ( www.loper-os.org ) --

asciilifeform: --- (C) 2017 Stanislav Datskovskiy ( www.loper-os.org ) --

asciilifeform: in raw diff ,

asciilifeform: and oh forfuckssake, apparently asciilifeform cannot change any of the '2017' to '2018' in ffa , because :

asciilifeform: fwiw ffa is endian-insensitive .

asciilifeform still writing ch9, which is actually about this.

asciilifeform: though to be fair i have not tested sparc , so possibly there is a third...

asciilifeform: phf: phunphakt : ppc is one of the two currently-produced archs ( the other being arm ) that doesn't have constant-time MUL .

asciilifeform: right

asciilifeform: http://btcbase.org/log/2014-10-14#872853 , but definitely had moar detailed thread prior ☝︎

asciilifeform: possibly was in comment thread.

asciilifeform: ( www )

asciilifeform: suggests that it was in trilema

asciilifeform: http://btcbase.org/log/2014-10-15#876019 << possibly prior to this mark ☝︎

asciilifeform: lessee..

asciilifeform: i definitely recall this

asciilifeform: oh hm

asciilifeform: was this the 'what if nothing could decay' thing ?

asciilifeform: mircea_popescu: what was the context there ?

asciilifeform: if done correctly.

asciilifeform: even on somewhat aged iron

asciilifeform: very often 'too expensive' turns out -- isn't

asciilifeform: ffa for instance is result of 'why not see what constant-time costs'

asciilifeform: hey why not see what bitwise algo costs

asciilifeform: yer 'supporting idiocy' when you uncrate and plug in x86, lol

asciilifeform: ( i dislike, on aesthetic grounds, but that's possibly just me )

asciilifeform: mircea_popescu dislikes the 'we know the constants' endianism, flip if machine's is opposite' traditional pill ?

asciilifeform: ( luggage does on occasion end up in /dev/null )

asciilifeform: supposing mircea_popescu is happy to risk a coupla btc of crate 'in 1 basket'

asciilifeform: aha

asciilifeform: what's a couplea moar.

asciilifeform: from BingoBoingo's photos i picture a slightly shabbier buenos aires . but yes, i already have 3 crates that want to go

asciilifeform: and even then has nonzero chance of 'hey this is the 3rd time this evil-looking gaijin is bringing 3 'personal' boxen'

asciilifeform: i suspect that the suitcase method only beats the 22% vig if 2-3 machines at a time are moved

asciilifeform: BingoBoingo: got it

asciilifeform: oh hey asciilifeform's algo

asciilifeform: panda tears for wiper tank.

asciilifeform: that's for radiator !

asciilifeform: lol

asciilifeform: didja inflate with argon too

asciilifeform: neato.

asciilifeform: where mircea_popescu gave a picture, 'i'ma buy a 6th item in the place that has the correct 1-5 already, and even for moar dough'

asciilifeform: iirc this was discussed in old thread

asciilifeform: and why not go. i'd actually much like to bring the three s.nsa boxen over with own hands

asciilifeform: supposing that it is actually necessary, and those heathen outfits would not ship straight to bb