UmbralRaptor 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> … one of the other grad students just compared me to nomal O_o | <ferram4> I shall beat my problems to death with an engineer.
<bofh>
UmbralRaptor: so like, now I'm curious, what sort of Mars orbit would one land in were you to trip and fall off of Phobos?
<bofh>
or would one get Roche'd?
<egg|zzz|egg>
bofh: you're not gravitationally bound (well if you are I have questions), so the Roche limit isn't very relevant to you is it?
<bofh>
egg|zzz|egg: well presumably you'd be gravitationally bound to Mars in a low-Martian-orbit
<bofh>
the question is how stable is it
<egg|zzz|egg>
bofh: yes, but the Roche limit is about you being gravitationally bound to yourself
<egg|zzz|egg>
and either electromagnetism dominates or, well what are you exactly
<bofh>
okay point. I guess you are smaller than, like, Enceladus
<egg|zzz|egg>
bofh: so you'd probably end up in an orbit mostly similar to phobos, but the question is whether (and how) you'd interact with Phobos again
<bofh>
well I guess that depends on both what your normal to Phobos is and your insertion velocity?
<egg|zzz|egg>
bofh: ... you should experiment with RealSolarSystem+Principia :-p
e_14159 has quit [Ping timeout: 200 seconds]
<bofh>
good idea :P
<bofh>
sec
<egg|zzz|egg>
(Principia is actually fairly inaccurate on phobos's orbit, we'd need all the 2nd order harmonics, not just J2~C20
<bofh>
jesus christ the --drg-hsm-error flag to hdparm
<bofh>
--drq-hsm-error
<bofh>
VERY DANGEROUS, DON'T EVEN THINK ABOUT USING IT. This option causes hdparm to issue an IDENTIFY command to the kernel, but incorrectly marked as a
<bofh>
"non-data" command. This results in the drive being left with its DataReQust(DRQ) line "stuck" high. This confuses the kernel drivers, and may crash
<bofh>
the system immediately with massive data loss. The option exists to help in testing and fortifying the kernel against similar real-world drive mal‐
<bofh>
functions. VERY DANGEROUS, DO NOT USE!!
<bofh>
actually the number of hdparm options marked "VERY DANGEROUS DO NOT EVER USE" is kinda terrifying
<kmath>
<QuantaMagazine> ICYMI: A French mathematician has completed the classification of all convex pentagons that tile the plane.… https://t.co/tXIikMVmfg
<egg|nomz|egg>
bofh: ^
* UmbralRaptor
was half-expecting to see "Robin Leroy" in that article for some reason.
* UmbralRaptor
stabs the headers on that site with Firefox's reading mode.
<bofh>
oh god, amazing.
<egg|nomz|egg>
(via @spun_off)
<lamont>
egg: i may need to get J2 out of principia sooner rather than later
<egg|nomz|egg>
hmm
<lamont>
my 185x185 orbits launched from KSC are off
egg|nomz|egg is now known as egg
<egg>
lamont: huh, by how much
<lamont>
182x188 or something on that order
<lamont>
not the fake orbit
<lamont>
i need to manually rub some J2 on it today and see if that fixes it first, which i think should be reasonably simple
<egg>
lamont: yeah that seems like the best solution for now
<egg>
!wpn Sarbian
* Qboid
gives Sarbian a sonic screwdriver
<egg>
!acr -add:adj torx
<Qboid>
egg: I added the explanation for this acronym.
<egg>
uh
<egg>
!wpn -add:adj torx
<Qboid>
egg: Adjective added!
<egg>
!acr -remove:adj
<Qboid>
egg: I removed the explanation for adj
egg|cell|egg has quit [Quit: Bye]
<egg>
bofh: I wonder whether I should TeX some documentation for that trajectory downsampling method or whether the code + comments are clear enough
* egg
stares at incoming ネコバス
<bofh>
egg: I think the comments are clear but the code is a bit tricky at times, it's worth TeXing up if you have time tho since the technique is solid and interesting in itsself.
<egg>
bofh: what do you mean the code is a bit tricky, it just has an irreducible goto into a nested loop :-p
<kmath>
<rygorous> @LiaSae @FioraAeterna @stephentyrone Yeah, pow by 220 is one giant round-off error amplifier: as noted elsewhere, (2046/2048)^220 = 0.8981.
* egg
likes that Turing quote on "ill-conditioned"
<egg>
"The expression 'ill-conditioned' is sometimes used as a term of abuse applicable to equations, but it seems most often to carry a meaning somewhat similar to that defined below."
<egg>
Fiora: I'm having a hard time looking it up, so, what's with the weird range?
<Fiora>
so
<egg>
Fiora: or rather, what horror happens outside
<Fiora>
x * y = x * (1 / y), allowed transformation in fast-math GL/et
<Fiora>
*etc
<egg>
hm
<Fiora>
so
<egg>
uh
<Fiora>
suppose we're doing 2^126 / 2^127.
<egg>
denormals only close to 0, not close to infinity?
<Fiora>
1 / 2^127 is a denormal.
<Fiora>
flush to zero.
<Fiora>
2^126 * (1 / 2^127) is zero.
<egg>
ooh flushing denormals
<Fiora>
thus, 2^126 / 2^127 is zero.
<Fiora>
infinite ulp error ;-)
<egg>
Fiora: hm, yeah
<egg>
Fiora: flushing denormals is shitty, but I think that decomposition would do the same even if you didn't flush them? since 2^(-149) would invert to infinity
<Fiora>
yes, but 2^-149 doesn't exist in GL.
<egg>
Fiora: aha, so at least you don't have to worry about denormal inputs, just normal inputs that catastrophically fail because of the flushing of denormals :-p
TonyC has quit [Ping timeout: 183 seconds]
regex has quit [Remote host closed the connection]
<Fiora>
yup. it's an internal issue caused by the fact that normal inputs require denormal intermediates
<egg>
Fiora: also, infinite relative error sure, but I thought by definition ULP distances were always bounded?
<Fiora>
well, they SHOULD be bounded.
<Fiora>
you can encounter similar issues when using newton-raphson (internally) in the computation of things
<egg>
Fiora: no I mean bounded as in I thought the distance between 0 and infinity was the number of representable values in between, not the (very obviously infinite) relative error
<Fiora>
ex: in our expansion of sqrt(x), for non-fast-math, we do N-R to refine to precise
<Fiora>
but if the input is < 2^-63 we have to pre-multiply by 2^32
<Fiora>
so that we don't flush denormals to zero inside of N-R
<Fiora>
and... yeah probably, to what you said? "infinite ulp" is kinda a casualism
<egg>
(tbh I've never counted anything in ULPs in numerics class, only relative errors)