178400+ entries in 0.107s
mircea_popescu: at
the very least should give some indication what
to look at closer
a111: Logged on 2017-07-03 05:50 mircea_popescu: ok, so for every string of FG bits called M, do hashing of M, M+1, M+2, M with first bit flipped, M with first
two bits flipped, inverse of M.
that's 6 items.
mircea_popescu: "what's
the difference between whimsical and whipsical ?" "the welts."
mircea_popescu: people have
this uncanny ability
to describe abstracts
they do not in
the least comprehend.
a111: Logged on 2017-07-03 14:54 asciilifeform: whereas for item
that ~you wrote~, you already have
the kind of picture
that a sampling profiler is giving you -- in your head
mircea_popescu: you're not gonna be remembered as "that guy who spent
the whole
time
touristing" are you.
a111: Logged on 2017-07-03 14:38 phf came back from
the woods again
phf: if you quicklisp bulk of your code, you're still basically in honeywagon
territory
phf: i
think maybe it's worthwhile as an auxiliary exploration. you can run it and it will either confirm what you already understood, but sometimes it'll correct your understanding. either way you have
to understand how sampling profiler works (particularly when it comes
to sampling freq),
to use it effectively
phf: you can do sampling profiler
to get some idea of what your mainloop looks like, but in
the code of
that size you should be able
to reason about it. you
then ensure
that you have best possible
type for your
task, and
type annotate
the mainloop functions.
type annotation for speed is
tricky and unless you know both
the standard and your implementation in and out, you have
to look at
the assemble of
the resulting function
phf: i
think ascii already made
that point,
that if you're profiling lisp with
the vm startup,
then you should also profile c machine from boot
time. at
the very least
the vm should be warmed up by loading all
the dependencies into
the core, doing save-lisp on it, and
then making sure
that your foo.lisp has an up
to date fasl. inside lisp
though
to achieve
the optimizations you run variants of your function inside (time ...) until you bring it within
the ra
☟︎ phf: it's a special
thread alright.
phf came back from
the woods again
☟︎ mircea_popescu: this is
turning into some pretty hardcore computering, 8 versions per M, esp in
the 64kb / 2kb format
mircea_popescu: add
to
that 2 instances of a single randomly located bit flipped.
mircea_popescu: ok, so for every string of FG bits called M, do hashing of M, M+1, M+2, M with first bit flipped, M with first
two bits flipped, inverse of M.
that's 6 items.
☟︎ ben_vulpes: mircea_popescu: keep 'em coming, experimental design in
this space is not my strength
ben_vulpes: going
to look at output distribution
too
ben_vulpes: uniform strings; still setting all
this machinery up
mircea_popescu: ben_vulpes you should prolly also capture
the M/H pairs
ben_vulpes: also, was wrong last night, go impl still
trouncing
ben_vulpes: 65536/2048 now at wall clock
time of 5.5 hrs
ben_vulpes: mircea_popescu: i'll letcha know after i rerun and capture
that
lobbes: I gotta say, I'm experiencing
the whole "WoT make decisions easier"
thing in practice now. I.e. I know hanbot is good for it, and since I
trust danielpbarron's rating of Birdman, I also
trust
that he,
too, will be good for it
lobbes: mircea_popescu, hehe. I was
thinking of implementing
that once n00bs start using
the
thing. Make it Wot-based at some level; disabling auction service for people lower down in WoT once delinquent x amount of
time
mircea_popescu: lobbes are you going
to not let delinquents auction
till
they pay ? :D
mircea_popescu: that'd be
the other point of interest, "how big does S get"
ben_vulpes: in other news, 65536 byte message, 256 bit hash
took some 2.47 hrs
ben_vulpes: i will probably just implement
the slice
that i actually need
ben_vulpes: yes
this refrain does ring a few bells
a111: Logged on 2017-07-02 15:36 asciilifeform: btw ben_vulpes your mphash seems
to use some shitlibrary
that 1) i dun have 2) won't install via quicklisp
a111: Logged on 2015-03-31 02:52 asciilifeform: i will now
take
the liberty of calling
the hypothetical machine 'freyacopter'
mircea_popescu: this incidentally is a very undiscussed
topic, suspiciously. what is
the effect of generating BOTH
the padding for a message and
the key
that's used on
the same... historical debian / kochatron /etcv
mircea_popescu: asciilifeform it's well known, i had conversations re
this in 2012. prolly can fish out quote.
mircea_popescu: anyway -- if you're making both
the payload and
the hash on
the same machine...
mircea_popescu: this aside ; i'm kinda loath
to mix rsa in both padding and encryption. for all anyone knows
they resonate
mircea_popescu: this is computationally worse
than mpfhf
though. imagine, nextprime(todays log).
mircea_popescu: "does it at least support unicode ?" "no. you may not
teach reading
to arabs, it's a sin."
mircea_popescu: large part of usgization of normal processes. "oh, does your msg app support VIDEO ?" bitch, i don't want
to watch overweight dudes with
their cats braying into a mic.
mircea_popescu: or other
things. and if you don't must, you FUCKING DIDN'T NEED
THE GB.
mircea_popescu: basically,
there's a large number of perfectly useful applications for 64kb capable hash. and upon examination it may be discovered
that larger sizes not supported is not a bug
mircea_popescu: and in other lulz, #trilema OODA loop : "hey, make an X" "what kind an X ?" "JUST MAKE AN X!" *picks arbitrary params a
through w, makes such an X. "here you go" "o cool, now set b
to
this and q
to
that, link l
to k and make c double d" "here." "holy shit
the submarine now flies! and lays eggs! good job man!"
a111: Logged on 2016-12-24 01:02 asciilifeform: incidentally, since (to borrow
the lament of
turing's school headmaster) 'the room already stinks of mathematics', i'll share a
tidbit
that i promised folx some half year ago and promptly forgot: