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.
<SnoopJeDi>
awang, the plain-language version is that Integrating Things Is Hard and if you aren't careful you can fail to conserve energy in a system (a lot of egg|anbo|egg's numerical eggsperiments are concerned with this)
<awang>
The mathematics gradient between the Principia codebase and my head may prove to be too much for that to work :(
<SnoopJeDi>
The Boris pusher splits up the way you move a particle in electrodynamic simulations in a way that avoids this. The n+1/2, n-1/2 business is describing "half" steps
<awang>
Although since I have 50% of the things needed to learn graduate math, I suppose it's worth a shot
<SnoopJeDi>
it's a *little* like something called the leapfrog method
<SnoopJeDi>
numerics are...hard.
<awang>
....Apparently o_O
<awang>
That would also explain why a FEA tool at work likes to crash after changing an input by a small amount, I guess?
<awang>
FEA involves integration?
<SnoopJeDi>
uhhhhh
<SnoopJeDi>
sorta? in the sense that you usually build your problems in integral form
<SnoopJeDi>
most of the machinery that makes FEA go is more akin to variational calulus, though
<SnoopJeDi>
awang, are you familiar with Newton's method?
<awang>
Sort of
<awang>
Been a while since I did anything calculus-related
<SnoopJeDi>
if you're comfortable with the "update the answer until the answer is good enough" sort of scheme, that's all you *really* have to know in principle. FEA is a very-high-dimensional analogue (and indeed, many solvers in commercial products use things that look a lot like Newton's method!)
<awang>
egg|anbo|egg: Where'd the "Leapfrog" part come from?
<awang>
Ah
<awang>
That's... surprisingly simple
<SnoopJeDi>
from a linear algebra standpoint, you're trying to find the closest neighbor to a thing that lives in infinite-dimensional space (the "true" solution)
<SnoopJeDi>
where the neighbor lives in finite-dimensional space (your mesh)
<SnoopJeDi>
which is incidentally why "mesh it right" is like 99% of the battle haha
<awang>
Makes sense
<egg|anbo|egg>
awang: trying to describe the fact that your velocities and positions are offset half a step from each other, so advancing one leaps it over the other
<awang>
That would also explain some of the conversations around the office talking about meshes
<egg|anbo|egg>
(afaict nobody called Leapfrog invented it)
<SnoopJeDi>
(as opposed to updating the positions and velocities simultaneously)
<egg|anbo|egg>
(but then a lot of people invented it, so you never know)
<SnoopJeDi>
oh yeah, meshes are the bane of all FEA
<SnoopJeDi>
even if the mesh doesn't make a problem unsolveable, a bad choice can make the problem poorly conditioned and it will fail to converge
<awang>
egg|anbo|egg: Ah, that name makes sense
<SnoopJeDi>
if you're curious about it, the FEniCS project has a whole book that I believe has some pretty good info on FEM as well as how to use their Python codebase to solve problems
<SnoopJeDi>
I learned mine out of a monograph by Peter Monk, "Finite Element Methods for Maxwell's Equations," which is kind of a (very good!) hodgepodge of functional analysis, electrodynamics, numerical analysis, and a few other odds and ends
awang has quit [Ping timeout: 186 seconds]
awang has joined #kspacademia
<awang>
SnoopJeDi: I'll definitely check that out
e_14159 has quit [Ping timeout: 198 seconds]
e_14159 has joined #kspacademia
icefire has quit [Quit: Leaving]
icefire has joined #kspacademia
icefire has quit [Remote host closed the connection]
icefire has joined #kspacademia
icefire has quit [Client Quit]
icefire has joined #kspacademia
icefire has quit [Remote host closed the connection]
icefire has joined #kspacademia
icefire has quit [Remote host closed the connection]
icefire has joined #kspacademia
icefire has quit [Remote host closed the connection]
icefire has joined #kspacademia
icefire has quit [Quit: Leaving]
icefire has joined #kspacademia
egg|anbo|egg has quit [Read error: Connection reset by peer]
<kmath>
<diodelass> @bofh453 the amusing thing is that these days I can get $100+ worth of electronic parts on the deparment's budget w… https://t.co/p8jqoSPyNQ
Majiir has quit [Ping timeout: 186 seconds]
icefire has quit [Read error: Connection reset by peer]
Snoozee has joined #kspacademia
Snoozee is now known as Majiir
awang has quit [Ping timeout: 207 seconds]
ferram4_ has joined #kspacademia
ferram4 has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
ferram4 has joined #kspacademia
ferram4_ has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
<kmath>
<whitequark> *stares at $300 in S3 bills* this is fine
<UmbralRaptor>
Hah
egg|laptop|egg_ has joined #kspacademia
egg|laptop|egg has quit [Ping timeout: 180 seconds]
StCypher has joined #kspacademia
StCipher has quit [Ping timeout: 186 seconds]
<UmbralRaptor>
So, you know how I took 2 qualifying exams on January 17 and 18? Apparently they haven't been graded, and probably won't for another week or 2. So, I should find out if I actually passed Classical and Quantum sometime in mid February.
* UmbralRaptor
🔪 🗡🔪
<SnoopJeDi>
>:|
<soundnfury>
UmbralRaptor: so, currently you (|passed> + |failed>)/√2 quals?
<UmbralRaptor>
Yes
<soundnfury>
don't worry, you'll decohere when the results arrive
<soundnfury>
or possibly incohere, whatever
<soundnfury>
(but you won't collapse ;)
<UmbralRaptor>
Probably
egg|laptop|egg_ is now known as egg|laptop|egg
icefire has joined #kspacademia
egg|laptop|egg has quit [Ping timeout: 180 seconds]
egg|laptop|egg has joined #kspacademia
<egg|laptop|egg>
bofh: okay, today's episode of "principia development": a benchmark that, if run with profiling, causes the machine to powercycle
<egg|laptop|egg>
it's deterministic
<egg|laptop|egg>
I have no idea
<APlayer>
egg: It is so overwhelmed from the BEAUTIFUL MATHS that it briefly falls unconscious
<APlayer>
On a more serious not, the single most foolproof way of crash debugging I know is manually tracking the program flow, installing breakpoints on the way and eventually tracking it down to a single routine, block of code and line.
<APlayer>
Might get impractically long in complex flowing programs, though
<UmbralRaptor>
Work the delta function-related problems in Griffiths and Jackson.
<UmbralRaptor>
Optional: Find an old graduate E&M textbook. It probably does not introduce the delta function, but it covers topics similar to that covered in both Griffiths and Jackson. How do they handle delta function related problems?
tawny- has quit [Ping timeout: 383 seconds]
<UmbralRaptor>
^how is this a single homework problem?
<UmbralRaptor>
egg|anbo|egg: coronagraphs are magic.
<egg|laptop|egg>
bofh: a spectre is haunting Principia: the spectre of speculative execution.
<SnoopJeDi>
egg|laptop|egg, are you familiar with Floquet theory at all? I'm scratching my head a bit at why the transformation is valid for Hill's equation when the varying coefficient κ(s) is nonperiodic
<UmbralRaptor>
egg|anbo|egg: will building Principia on my MBA cover my desk in corium?
<egg|laptop|egg>
UmbralRaptor: unlikely
<egg|laptop|egg>
SnoopJeDi: never heard of it tbh
<SnoopJeDi>
I suspect if the coordinate s is on some bounded interval, you do the "normal" trick of copy-pasting κ(s) infinitely to make it periodic and discard the solution outside the interval anyway
<SnoopJeDi>
ah okay, figured I'd try my luck :P
<egg|laptop|egg>
UmbralRaptor: but you should build principia!
<egg|laptop|egg>
SnoopJeDi: try bofh
<SnoopJeDi>
the book by the Very Smart Man says the solutions are valid even in the nonperiodic case, but I'm bothered by why and my advisor is being a dismissive dick about it
<awang>
Wonder what about the sampling code causes Windows 7 to reboot without any logs or anything
egg|phone|egg has joined #kspacademia
egg|laptop|egg has quit [Ping timeout: 180 seconds]
<bofh>
on today's episode of hell, compiling libtiff for windows-i386.
<awang>
bofh: Why?
APlayer has quit [Ping timeout: 182 seconds]
<UmbralRaptor>
Party like it's 1998?
<UmbralRaptor>
Or I suppose 2008. *stares at Urvogel*
<awang>
Urvogel?
<UmbralRaptor>
Context, an old netbook I have (Intel U2500, a Core1Duo).