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.
<egg>
bofh: ... wait idea, if I just use the 32x32 mul I still have plenty of bits to do the integer approximation stuff
<egg>
(probably makes the error analysis messy though)
<SnoopJeDi>
UmbralRaptop, wait so are your data inhomogeneous? ndarray has to be one dtype, or you have to give it a tuple of dtypes for tabular mixed data
<SnoopJeDi>
if you're doing that though, you *probably* want astropy or pandas?
<SnoopJeDi>
numpy's "named" columns are weird to me, it feels like a bolt-on feature
<UmbralRaptop>
SnoopJeDi: some columns have floats, some have strings. The contents of any one column are homogeneous.
<SnoopJeDi>
idk if astropy lets you construct Tables from arbitrary data, though
<SnoopJeDi>
probably does though because it's meant to be a cleaner DataFrame for astronomers
<UmbralRaptop>
Probably.
* UmbralRaptop
stares at the FITS format, and how it explicitly references tapes and cards. >_> <_<
<iximeow>
ah, fits
<iximeow>
i'm going to have to write things to deal with fits, i fear
<SnoopJeDi>
oh if it's FITS you *definitely* want astropy!
<SnoopJeDi>
astrop.io.fits \o/
<SnoopJeDi>
astropy*
<SnoopJeDi>
unless you just mean to point out FITS as a not-great example of technological baggage :P
* SnoopJeDi
stares at OPAL, which outputs to HDF5...and stores every step of Runge-Kutta on its own Group
<SnoopJeDi>
if you wanna do time-series anything glhf
<UmbralRaptop>
I'm being silly about FITS, it's a CSV
<SnoopJeDi>
mkay
UmbralRaptor has joined #kspacademia
UmbralRaptop has quit [Ping timeout: 182 seconds]
e_14159 has quit [Ping timeout: 186 seconds]
e_14159 has joined #kspacademia
<egg>
bofh: hm so to check for infinities, denormals, etc. the SSE4.1 test instructions seem nice but currently we're aiming for SSE3 (so as to not introduce restrictions beyond x86-64)
<egg|mobile|egg>
RT if you have any questions or concerns please visit the plug-in settings to determine how attachments are handled the situation and I will be nice to see you soon and I will be nice to see if you have any questions or concerns please...
<egg|mobile|egg>
Seems to diverge eventually
egg|phone|egg has joined #kspacademia
egg|mobile|egg has quit [Read error: Connection reset by peer]
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
egg|phone|egg has joined #kspacademia
egg|cell|egg has quit [Ping timeout: 186 seconds]
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|phone|egg has joined #kspacademia
egg|mobile|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|cell|egg has quit [Ping timeout: 186 seconds]
egg|mobile|egg has quit [Ping timeout: 186 seconds]
egg|phone|egg has joined #kspacademia
awang has joined #kspacademia
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Ping timeout: 182 seconds]
awang has quit [Remote host closed the connection]
egg|phone|egg has joined #kspacademia
egg|cell|egg has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
<egg>
!wpn whitequark
* Qboid
gives whitequark a sparse nose
<egg>
bofh: tbh I'm surprised that this is happening already
<egg>
bofh: do you have something that's neither a sandy nor a skylake to compare for other architectures
<egg>
!wpn bofh
* Qboid
gives bofh a stretchy missile-like power supply
<egg>
!wpn rqou
* Qboid
gives rqou a giant Spivak
<whitequark>
i have a kaby lake and a coffee lake
<whitequark>
and a uh
<egg>
"coffee lake" sounds fun
<whitequark>
oh no that one is a skylake
<egg>
whitequark: tbh building principia is interesting, building that odd branch where I've been doing my cbrt benchmarks is probably too much of a mess
<whitequark>
egg: just build it for linux?
<whitequark>
and gimme a binary
<egg>
not sure I've ever built the benchmarks on linux :-p
<egg>
whitequark: also I guess the really interesting result would be to see on older architectures whether the polynomial was significantly faster than the padé, both sandy and skylake are very close https://twitter.com/eggleroy/status/988114872616484866
* egg
afknomz
egg is now known as egg|nomz|egg
<UmbralRaptop>
In principle, I have a Yonah lying around.
<SnoopJeDi>
oh wow, one of our profs was awarded a *second* Landau medal this year
<bofh>
SnoopJeDi: NICE
<bofh>
egg|nomz|egg: I have a Haswell & a Nehalem handy
<SnoopJeDi>
he's apparently the only person who's received it a second time, but also wicked smaht
<bofh>
egg|nomz|egg: few mins
<bofh>
SnoopJeDi: very nice
UmbralRaptop has quit [Quit: Bye]
UmbralRaptop has joined #kspacademia
UmbralRaptor has joined #kspacademia
UmbralRaptor has quit [Remote host closed the connection]
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptop has joined #kspacademia
UmbralRaptop has quit [Read error: Connection reset by peer]
<bofh>
egg: wasn't he from like Macedon, so like 400BCE or later?
<bofh>
how the heck was his name in a document from Ur?
<UmbralRaptor>
Vampires.
<egg>
bofh: the game is from Ur, the tablet is late
<rqou>
egg: not sure exactly what you're doing with benchmarking, but if you want i can run your stuff on a Diamondville N270 if you want some fun totally-different uArch
<bofh>
rqou: I mean I can also run this stuff on a ev68 DEC alpha :P
<bofh>
(also ARMv7-NEON & ARM64, but those are more standard)
<egg>
!wpn rqou
* Qboid
gives rqou a RF Ibuprofen
<UmbralRaptor>
Anyone got an M68k system lying around?
<UmbralRaptor>
!wpn egg
* Qboid
gives egg a xenon VAX
<rqou>
yes
<rqou>
but it's not set up to run any custom code at this point in time
<rqou>
also, not a traditional "computer"
<rqou>
actually now that i think about it i actually have quite a few
<SnoopJeDi>
mother of god this sentence: "The dynamics of acceleration of a high-current particle bunch is dominated by the evolution of its 6-D phase space through the lattice of magnets and accelerating cavities, including the self-action of space charge and wake fields from the same bunch and its neighbors."
<bofh>
uhm.
<bofh>
"dominated by"? I feel like that's describing the entirety of the interaction of the particle >_>
<SnoopJeDi>
my advisor is a really intuitive guy but he just spews when he writes something
<SnoopJeDi>
thankfully I knew not to get out ahead of him too far on this draft, and instead try to slowly infect his "what the fuck did I just read" with sensible prose
<bofh>
rofl
<bofh>
++
<SnoopJeDi>
bofh, I know where he's coming at it from, though. When we construct Hill's equation, we're truncating an expansion of the Hamiltonian
<SnoopJeDi>
and accelerator physics is largely a hodgepodge of sorta-perturbative sorta-bullshit techniques built up around understanding the other stuff in lieu of very robust analytical theories
<SnoopJeDi>
and/or "fuck it throw a million macroparticles down the bore and let's see what comes out the other side"
<SnoopJeDi>
the latter being largely what we've done because this machine is so hinky
<bofh>
Okay, true, but those Hamiltonian expansions often contain many terms whose contribution to the system's eigenvalues is negligible and which, if left in, lend the system to be not just non-analytically-soluble but also hell to approach computationally.
<bofh>
ROFL
<bofh>
I approve of that approach.
<SnoopJeDi>
yea so
<SnoopJeDi>
it turns out that with the Hamiltonians we care about
<SnoopJeDi>
there are actually proofs that they *arent* negligible
<SnoopJeDi>
a lot of the time we don't run into issues with that because of beam lifetimes due to [other stuff] though, so it's a hodgepodge like I said
<SnoopJeDi>
like yea I could keep this beam around without substantial loss for another 100k turns, but the emittance has gone to shit so the luminosity will suck, it's better to just beam-dump and get fresh beam
<egg>
bofh: hmm how do I write the overflow-preventing rescalings for my cbrt
<SnoopJeDi>
which is kind of an upsetting way to think about accelerators from the perspective of energy, but the wall-plug power you "throw away" is probably dwarfed by stuff like running the cryo and creature comforts like lighting and breathable air
<egg>
!wpn whitequark
* Qboid
gives whitequark a canonical asura which strongly resembles a crystal
<egg>
!seen котя
<Qboid>
egg: Value cannot be null.
<SnoopJeDi>
!seen котяegg
<Qboid>
SnoopJeDi: Value cannot be null.
<whitequark>
!wpn egg
* Qboid
gives egg a Balmer tesseract
<SnoopJeDi>
oh from the same draft: "A PYTHON shell has been written to access those field maps in OPAL's time domain simulation."