egg|nomz|egg changed the topic of #kspacademia to: https://gist.github.com/pdn4kd/164b9b85435d87afbec0c3a7e69d3e6d | Dogs are cats. Spiders are cat interferometers. | Космизм сегодня! | Document well, for tomorrow you may get mauled by a ネコバス. | <UmbralRaptor> egg|nomz|egg: generally if your eyes are dewing over, that's not the weather. | <ferram4> I shall beat my problems to death with an engineer.
hillexed has joined #kspacademia
hillexed has quit [Client Quit]
e_14159_ has joined #kspacademia
e_14159 has quit [Ping timeout: 202 seconds]
<rqou>
bofh: are there any general statements you can make about the numerical stability of algorithms using (ieee 754) floating point vs using fixed point?
<rqou>
more specifically, in this case (i believe) i want to implement quadratic optimization using the conjugate gradient method
<rqou>
what about my original question about fixed vs float and numerical stability?
<egg>
general statements are hard :-p
<egg>
floating point behaves nicely multiplicatively, not so for fixed
<rqou>
ok, how about in the specific case of " i want to implement quadratic optimization using the conjugate gradient method"
<egg>
rqou: try summoning bofh, I cannot into optimization
* egg
casts "summon bofh"
<rqou>
i tried summoning bofh earlier :P
<egg>
oh right, that said bofh but pinged me because numerics :D
<egg>
rqou: it's a bit early for bofh though, considering that bofh is currently in europe
<bofh>
egg: indeed, seeing as I just woke up
<egg>
:D
<egg>
bofh: besides rqou's question, double and topicquotes :-p
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
<egg>
!wpn bofh
* Qboid
gives bofh a radial cosmonaut with a vibrator attachment
<egg>
!wpn rqou
* Qboid
gives rqou a tesseral COME FROM
<egg>
!wpn whitequark
* Qboid
gives whitequark a symplectic Qboid
<whitequark>
uh
<bofh>
rofl
<egg>
a Qboid that preserves the symplectic form? that almost makes sense
tawny- has quit [Ping timeout: 194 seconds]
egg|cell|egg has quit [Ping timeout: 383 seconds]
<egg>
bofh: aargh the flies are annoying
<bofh>
egg: which files?
<egg>
bofh: s/il/li/
<Qboid>
egg thinks bofh meant to say: egg: which flies?
<egg>
at ANBO
<egg>
the dipterans
<bofh>
ahh
<bofh>
WNGANT~1.EXE?
* egg
stabs KSP
<egg>
so, since the double-double-ish mechanism for position and velocity are separate and independent, of course updating position and velocity following a change of SoI happen at completely different times...
<egg>
stabbity stab
<bofh>
that reminds me, double-double & optimization stuff
<bofh>
l
<bofh>
gods my feet hurt
<egg>
bofh: have you been walking around again
<bofh>
like this entire past week has been nonstop walking :p
<bofh>
finally like have a moment to settle doen a bit this afternoon
<egg>
bofh: still nothing on the котя visa? :-/
<bofh>
negatory, but I'm planning to be back in EU in like 2-4 months anyway :P
<bofh>
"This paper explores the physics of the what-if question "what if the entire Earth was instantaneously replaced with an equal volume of closely packed, but uncompressed blueberries?""
<APlayer>
I expected it to have been published on April the 1st, TBH
<egg>
bofh: the last sentence is fun
<APlayer>
"The effect is
<APlayer>
basically the worst earthquake ever"
<APlayer>
Well, that paper was a surprisingly good read. Thanks!
awang has joined #kspacademia
<egg>
!wpn Ellied
* Qboid
gives Ellied a Q-switched hbox
egg|phone|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
<bofh>
rqou: okay, my first question is: if you're trying to solve QPs, why are you using conjugate gradient anyhow? you *have* a Hessian after all, might as well use it.
<bofh>
that being said, fundamentally what you're implementing at its core when solving a QP is a line search and a POSV (i.e. solving a linear system, almost certainly +ve semidefinite). implementing that using fixed point is basically rescaling hell (admittedly, it is *highly predictable* rescaling hell, so it's at least feasible, just extremely tedious and nasty), and not an improvement over IEEE754 or
<bofh>
IEEE754-FTZ+DAZ.
<bofh>
curious as to why this came up anyhow, generally you fall back to fixed-point algos when working with limited resources or an FPU-less arch only, pretty much.
<kmath>
<sjs917> Oh but wait! It is found in G dwarfs - once Gaia DR1 came out, could filter our subgaints and the period bimodality popped out. #CoolStars20
<egg>
bofh: speaking of which, the cat doesn't seem to have meowed much at the door of the bedroom with the cat this week, did it do that while you were there? It did that back in may i think, and certainly in february
<egg>
or maybe i've just slept more than usual
<egg>
whitequark: btw did bofh send you that picture of a three-phase cable splitting into two two-conductor cables :-p
<bofh>
egg: it did it a bunch, honestly, mostly at 5AM or so, but less so than you made it sound like it would.
<egg>
whitequark: bofh: pair production in a 3-phase electrical installation,
<whitequark>
lol
* iximeow
looks at backscroll
<bofh>
egg: LOL
<egg>
bofh: i mean, one of those cables is 2 phases, you could do something with that if you wanted 380 V ungrounded for some inane reason, the phase + ground cable makes less sense...
<bofh>
egg: I mean you can always bypass the RCD and use the remaining phase as an earth return line...
<egg|phone|egg>
Well no, everything is behind the differential breaker :-p
<bofh>
yeah but that's nothing a bit of AWG#12 can't solve if you're comfortable with working on live circuits (and if you're intentionally bypassing a differential breaker, then why not? :P).
<egg|phone|egg>
Okay but then that's a different situation from the one in the picture, which is three phases behind breaker and fuses
<egg|phone|egg>
Also pretty sure doing anything behind the breaker is not regarded as cricket in the same way as doing things behind the counter: the breaker is also under seal and serves to limit your power too
e_14159_ has quit [Ping timeout: 202 seconds]
e_14159 has joined #kspacademia
<bofh>
egg|phone|egg: yeah, that makes sense.
pthag has quit [Remote host closed the connection]