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.
<UmbralRaptor> yay
tawny has quit [Quit: 「Roundabout」 - To Be Continued]
UmbralRaptor has quit [Remote host closed the connection]
UmbralRaptop has joined #kspacademia
<egg> !wpn bofh
* Qboid gives bofh a zonal Poké Ball with a gegenschein attachment
* UmbralRaptop approves of this coffee based solely on the name. https://photos.app.goo.gl/MSOU1y3iymnZGy8e2
e_14159 has quit [Ping timeout: 190 seconds]
<egg> Gqruda?
<UmbralRaptop> Garuda!
e_14159 has joined #kspacademia
<egg> bofh: I wonder whether I need more silly notation,
tawny has joined #kspacademia
<egg> bofh: so https://github.com/simonbyrne/apple-libm/blob/master/Source/Intel/cbrtf.s avoids divisions both floating and integer so I guess it might be faster than the approach detailed in Kahan's paper?
<egg> bofh: no idea what the error is in double precision though, is it as good as the 0,59 ULPs that Kahan has for double
<bofh> egg: so I solve that problem by instead computing the reciprocal of cbrt(x), squaring it and multiplying it by the original number, taking care to not overflow.
<bofh> (also you don't do any integer division there. it's division by a constant, so it's a multiply and a shift, both extremely fast)
<egg|cell|egg> Ah right
<egg|cell|egg> Hmm
<egg|cell|egg> Still no answer from the cat on that matter?
<bofh> I'm not sure if I handle all edge cases or how well it does compared to Kahan's paper, I need to do formal analysis of it sometime.
<bofh> Said he'll get back soon, is a bit busy atm.
<egg|cell|egg> Ow
* egg|cell|egg pets the cat
<bofh> Like I'm *pretty sure* I think I got all of them (actually let me just iterate over all possible floats when I get back to a computer and see what the max ULP error is, lol).
<egg|cell|egg> How many doubles are there in [1,8[
<egg|cell|egg> The periodicity should help
<egg|cell|egg> Hmm still too many
<bofh> Even in [1,2) there's too many, really.
<bofh> SnoopJeDi: so one of the talks I went to was this one and holy fuck it was an amazing brisk overview of the field: http://calendars.illinois.edu/detail/2568?eventId=33301740 (I asked the speaker about posting slides online, he said he's not sure but will email them to me at least).
<bofh> (also https://arxiv.org/abs/1108.3140 has a lot of it but it says a lot that this paper was maybe *half* the talk and about a quarter of it has since been shown to not quite be the case)
armed_troop has quit [Quit: Bye]
<UmbralRaptop> It's amusing how DSS images of high proper motion stars look like they have chromatic abberation. Or are outright split into orange and blue.
armed_troop has joined #kspacademia
<UmbralRaptop> !wpn -add:wpn bolometer
<Qboid> UmbralRaptop: Weapon added!
<UmbralRaptop> !wpn -add:adj bolometric
<Qboid> UmbralRaptop: Adjective added!
ferram4 has quit [Read error: Connection reset by peer]
ferram4 has joined #kspacademia
* UmbralRaptop stares at https://twitter.com/bofh453/status/979938939376201728 and ponders: silly or worrying?
<kmath> <bofh453> Might as well. https://t.co/qyqSQyM1uZ
<UmbralRaptop> (for characters, that is)
<bofh> I mean, either is fine :P
<bofh> UmbralRaptop: I mean aren't all spacecraft imagers filtered cameras, and so your blue and your orange pictures are taken a fraction of a second apart and hence wom't combinenicely if your target has high proper motion?
<UmbralRaptop> bofh: Yes, though DSS was from ground based surveys that took a number of years,
<bofh> oh, then you get bonus atmospheric ""fun"", too.
<bofh> also sadly not familiar enough with that manga to answer re: that character, tho it seems like an interesting series in any case.
<UmbralRaptop> ah
<UmbralRaptop> Today in random observations: since its discovery in 1916, Barnard's star has moved ~17.6' (or, more than half the apparent diameter of the moon)
<bofh> that's considerably more than I'dve expected tbh.
<bofh> anyway feel free to try another char :P
* UmbralRaptop blames Fiora for thinking in terms of anime and manga.
tawny has quit [Quit: 「Roundabout」 - To Be Continued]
tawny- has joined #kspacademia
tawny- has quit [Ping timeout: 190 seconds]
<egg|cell|egg> Bofh: U+A66E
<egg> !wpn whitequark
* Qboid gives whitequark a doomed achromatic thesis-like bird
tawny- has joined #kspacademia
tawny- has quit [Ping timeout: 190 seconds]
<egg> !wpn bofh
* Qboid gives bofh a trained point/nova hybrid
<kmath> <whitequark> cursed chipset https://t.co/L8wo5LNWht ⏎ it's an M.2 USB board that has a PCIe ROOT COMPLEX ⏎ that's right. it can HOST… https://t.co/2RwqFlI7ls
<egg> bofh: have you looked at what happens to ε after some householdering
<egg> bofh: also I wonder whether it would be feasible to benchmark atlas polynomial against Kahan's method (for slow div, where you compute 1/cbrt)
* egg afk sandwichmaking
egg is now known as egg|tea|egg
<egg|tea|egg> !wpn bofh
* Qboid gives bofh a pagan pentagon
* UmbralRaptop a_x
<UmbralRaptop> !wpn egg|tea|egg
* Qboid gives egg|tea|egg a copy-on-write Euclidean monorepo which strongly resembles a penguin
egg|tea|egg is now known as egg
tawny- has joined #kspacademia
<egg> !wpn UmbralRaptop
* Qboid gives UmbralRaptop a titanium conflict
<UmbralRaptop> stabbity?
tawny- has quit [Ping timeout: 186 seconds]
<bofh> halp I forgot how 2 Ising spin glasses
<UmbralRaptop> uh
<egg> bofh: um
<egg> !wpn bofh
* Qboid gives bofh an oceanic summation
<egg> !wpn -add:adj Ising
<Qboid> egg: Adjective added!
<egg> !wpn bofh
* Qboid gives bofh a xi buffer overflow
<egg> !wpn
* Qboid gives egg a perturbation
* egg perturbs bofh
<egg> !wpn rqou
* Qboid gives rqou an immiscible vbox
<egg> (badness 10000)
tawny- has joined #kspacademia
* UmbralRaptop wonders if a bus vanished.
<UmbralRaptop> Okay, just late.
<egg> !wpn UmbralRaptop
* Qboid gives UmbralRaptop a clopen icosahedron
<egg> !wpn UmbralRaptop's bus
* Qboid gives UmbralRaptop's bus a turbofish
<egg> !wpn bofh
* Qboid gives bofh a commuting polygon
<UmbralRaptop> !wpn tawny-
* Qboid gives tawny- a prescriptive vector
<kmath> <NileMagazine> Nothing says “Easter” like two rabbit-headed demonic doorkeepers of the Underworld, holding rearing cobras. ⏎ This de… https://t.co/Rei39KdjS6
<egg> bofh: so the signed relative error after one rootn Newton always has the sign of n, but if it didn't have the sign of n beforehand it can grow quite a bit in absolute value?
<egg> bofh: that might make its effect on the maxima annoying to understand
<egg> bofh: hmmm, otoh Halley is monotone on the signed relative error
<bofh> UmbralRaptop: by Marisa you mean the Touhou char, just to be sure?
<egg> wait no not if it's very negative?
<bofh> egg: the signed relative error of Newton does not behave nicely, yeah. :/ I *think* Halley isn't exactly monotone but it's close enough.
<egg> halp I cannot into error analysis of householder methods
<bofh> (again, I'll be more useful once this conference is over tonight)
<bofh> yeah, their error analysis pretty consistently seems to focus on the absolute error in every paper I've read.
<egg> bofh: ow, conference on easter week-end? :-/
<bofh> egg: https://physics.illinois.edu/news/article/24594 I'm not complaining, the talks are fantastic
<egg> bofh: also in case you want to know what the recurrence for the signed relative error on 3rd order Householder is:
<egg> 2 n n 2
<egg> (1 + n) (1 + ε) (-3 + (-1 + n) ε) + (-1 + n) (3 + ε + n ε) + 2 (1 + ε) (3 + ε + 2 n ε)
<egg> -------------------------------------------------------------------------------------------
<egg> 2 2 n 2 n
<egg> 2 - 3 n + n + 4 (-1 + n ) (1 + ε) + (1 + n) (2 + n) (1 + ε)
<egg> bofh: uh fourth I mean
<bofh> That looks appropriately vile, I *think* you can simplify that a bit but otherwise that looks extremely familiar.
<egg> the one after Halley
<bofh> Yeah.
<bofh> Halley's is nice-ish.
<egg> yeah
<egg> for Halley it is:
<egg> n
<egg> 2 + ε + n ε + (1 + ε) (-2 + (-1 + n) ε)
<egg> n
<egg> ----------------------------------------
<egg> -1 + n + (1 + n) (1 + ε)
<bofh> Yep!
<egg> (also I think there is an expanding brain meme to be made about this method for communicating equations)
<egg> Thomas, whitequark, I would like to file a bug report about the stripping of spaces in the logs: https://logs.tmsp.io/kspacademia/2018-03-31#1522518163-1522518163;
<egg> ah it's also out-of-order
TonyC has quit [Ping timeout: 198 seconds]
<egg> bofh: does it look reasonably decent on your end, or is it inherently broken
<egg> (i.e. is it Qboid's fault or mine)
<whitequark> it looks OK on my end
<whitequark> so it's a problem with logs
<whitequark> now regarding them being out-of-order... that's actually a known issue
<whitequark> insufficient timestamp resolution
<whitequark> I think
<egg> hm
<whitequark> or might be something else idk lol
<whitequark> I wrote that code so long ago
<egg> patching that compatibly might be annoying?
<whitequark> I just know that it does happen if you paste enormous amounts of text
<egg> but yeah it does sound like a timestamp resolution thing
<UmbralRaptop> bofh: I do.
<egg> it happens a lot with Qboid's !wpns
<egg> or kmath
<egg> e.g. sometimes the kmath expansion of a tweet will appear before the linking of the tweet
<kmath> <whitequark> https://t.co/eCaQO6WEB0
<whitequark> it might also be because of threading
<kmath> <whitequark> https://t.co/eCaQO6WEB0
<egg> hm
<whitequark> I think the IRC thing I'm using spawns a thread per message
TonyC has joined #kspacademia
<whitequark> hm
<egg> ow
<whitequark> but, timestamps
<whitequark> so yeah
<whitequark> timestamp resolution
<egg> yeah
<egg> that seems hard to compatibly patch
<bofh> egg: the ascii art expressions look quite nice (everything aligns correctly, and the exponents are in the right spots).
<bofh> UmbralRaptop: excellent
<egg> bofh: OK so it's the logging
<whitequark> egg: not necessarily
<whitequark> the db migration will take a while
<whitequark> but that's about it
<egg> hm
<egg> whitequark: and there's a separate issue that *something* is stripping spaces
TonyC has quit [Ping timeout: 383 seconds]
<egg> . . . . . . . . . .
<egg> ...
<egg> yeah, stripping leading and consecutive spaces
<bofh> I think that might even be intentional. Stripping leading spaces at least sounds like an intentional thing.
<whitequark> nope
<egg> yeah it sounds very intentional and also a bad idea in retrospect
<whitequark> that's just how HTML works
<egg> oh.
<egg> right.
<egg> derp, yes
<egg> it's all there in the source
<whitequark> so you can add white-space: pre-wrap but then you need to strip the unnecessary whitespace from the view.
<whitequark> that's like a minute of work.
<egg> what do you mean strip the unnecessary whitespace
<whitequark> the view template adds some whitespace on its own
<egg> !wpn whitequark
* Qboid gives whitequark a hacked endpoint
<egg> bofh: wait I'm an idiot
<egg> bofh: we don't care whether the error changes monotonically
<egg> bofh: we only care about any extrema that the error recurrence may have, as those may create new extrema of the error
<egg> bofh: there are no such extrema for Newton and Halley, but there seem to be some for higher Householder methods; also Mathematica chokes on that
<egg> bofh: Kahan does mention that the higher methods are useful for getting those last few ULPs though so they're still interesting to study...
<egg> also need to reboot
egg is now known as egg|rebooting|egg
egg|rebooting|egg has quit [Read error: Connection reset by peer]
<kmath> <ADDiane> Moses was technically the first person to download files to his tablet from the cloud.
<bofh> brb, moar talks.
UmbralRaptop has quit [Quit: Bye]
egg has joined #kspacademia
UmbralRaptop has joined #kspacademia
<egg> !wpn UmbralRaptop
* Qboid gives UmbralRaptop a stabby hydrochlorofluorocarbon
<egg> grmbl someday I should really write myself a decent IRC client
<UmbralRaptop> Teflon knife!
UmbralRaptor has joined #kspacademia
UmbralRaptor has quit [Client Quit]
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptop has joined #kspacademia
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptop has joined #kspacademia
<kmath> <✔TheTimeCowboy> its egg day https://t.co/PXL18ST2Gq
<egg> me day
<UmbralRaptop> But will you use Python?
* UmbralRaptop was expecting this to go somewhere else. https://twitter.com/dankmaths/status/980138323506290694
<kmath> <dankmaths> ignis shows her smarts for #cutiesaturday...... https://t.co/GDYsThYV8e
<UmbralRaptop> A kg of steel just weighs a kilogram. But a kilogram of feathers means you have to carry around the guilt of what you did to those poor birds.
* egg pets UmbralRaptop
* UmbralRaptop chirps.
<Qboid> [#133] title: Preparations for version 13. | DO NOT MERGE this pull request until both of the following conditions are met:... | https://github.com/KSP-RO/RealSolarSystem/issues/133
<UmbralRaptop> hrm
<UmbralRaptop> Resource: liquid metallic hydrogen
<egg> !wpn UmbralRaptop
* Qboid gives UmbralRaptop a dye otter/MissingNo. hybrid
<egg> !wpn whitequark
* Qboid gives whitequark a triacontahedron notion
<egg> !wpn bofh
* Qboid gives bofh a thermionic snowball
UmbralRaptor has joined #kspacademia
UmbralRaptop has quit [Ping timeout: 190 seconds]
<egg> !wpn UmbralRaptor
* Qboid gives UmbralRaptor an eager explicit bomb
<egg> !wpn bofh
* Qboid gives bofh a canonical Лидов–古在 vertical (with ground plane)
<egg> UmbralRaptor: Лидов–古在!
* UmbralRaptor is perturbed into a solar collision!
* UmbralRaptor 🔪 red dwarfs in their lack of data.
<egg> cat!
<egg> !wpn котя
* Qboid gives котя a hyperactive ham radio
<egg> !wpn whitequark
* Qboid gives whitequark a katana
<kmath> <bofh453> + extremely enjoy wandering & travelling ⏎ + a fan of clever problem solving ⏎ + am basically nobody ⏎ - not vindictive ⏎ -… https://t.co/HNyeAPhY5I