egg changed the topic of #principia to: Logs: https://esper.irclog.whitequark.org/principia | <scott_manley> anyone that doubts the wisdom of retrograde bop needs to get the hell out | https://xkcd.com/323/ | <egg> calculating the influence of lamont on Pluto is a bit silly…
raptop has quit [Ping timeout: 189 seconds]
raptop has joined #principia
_whitelogger has joined #principia
<queqiao-> ⟨Starscraper⟩ Trying to get a Venus flyby is extremely difficult, it seems. Is there any tricks I'm missing for figuring this out? Do I launch at the Hohmann window, or go for a shorter transit time or a longer one? I also would guess I lead in front of Venus' orbit around the sun to do the pass? Does it involve a burn at the Venusian periapsis or is it just an unpowered flyby?
<queqiao-> ⟨Starscraper⟩ +that then returns to Earth
<queqiao-> ⟨Transparent_Elemental⟩ there isn't a single correct approach to answer most of these questions with either yes or no
<queqiao-> ⟨Transparent_Elemental⟩ https://www.theplanetstoday.com/galileo_flight_path.html
<queqiao-> ⟨Transparent_Elemental⟩ there isn't a single correct approach to answer most of these questions with either a yes or no
<queqiao-> ⟨Transparent_Elemental⟩ there isn't a single correct approach to answer most of these questions with either a yes or a no
<UmbralRaptop> hrm, a 0.85 au (i.e. hohmann) has a period of ~0.78 years
* UmbralRaptop wonders if there's a cursed near-hohmann flyby where you ultimately go around Sol 5 times and return to Earth 4 years after launch
<queqiao-> ⟨Starscraper⟩ I'm sure there is
<queqiao-> ⟨Starscraper⟩ * is. Just transfer to an orbit with exactly an .8 year-period and you should have that 5:4 resonance (ie, flyby Earth in 4 years).
<queqiao-> ⟨Starscraper⟩ * years, after doing 5 orbits of your own).
<queqiao-> ⟨Starscraper⟩ ⟪Transparent_Elemental⟫ there isn't a single correct approach […] ⮪ Dang.
<queqiao-> Well, it looks like, going off of this, you pass on the sun-facing side to transfer into an orbit with an ap above Earth to do a second-intersect transfer back to Earth.
<queqiao-> ⟨Transparent_Elemental⟩ I would think this should be the opposite, i. e. the shadow side
<UmbralRaptop> Either way, there are both the 5:4 and 4:3 resonances available
<queqiao-> ⟨Serina⟩ huh, I just updated to the latest principia version and the orbits are kinda.. like flashing?
<queqiao-> ⟨Starscraper⟩ ⟪Transparent_Elemental⟫ I would think this should be the […] ⮪ p. 150 - it is indeed the sunny side: https://ntrs.nasa.gov/citations/19790072165
<queqiao-> You need to raise your aphelion to be considerably far above the Earth's so that you slow down in your solar orbit enough for the Earth to "catch up", so it makes sense in that regard.
<queqiao-> Also it's a 400-day mission. I misunderstood other stuff I was reading earlier.
<queqiao-> ⟨Starscraper⟩ This document seems to have exactly what I was looking for...
<queqiao-> ⟨Starscraper⟩ Oh, excellent, thanks for letting me know, I'll take a look at that next time I have some time.
<queqiao-> ⟨PrinceZola ùwú⟩ how hard does principia hits on your RAM / frames?
<queqiao-> ⟨PrinceZola ùwú⟩ i only have 8 gigs...
<queqiao-> ⟨PrinceZola ùwú⟩ but i wanted to try it out
<queqiao-> ⟨Starcatcher⟩ Seems like RSS hits me more on the RAM.
<queqiao-> I recommend you to use Mem Reduct and KSP Allocation Fix, it helps reduce ram usage.
<queqiao-> ⟨Starcatcher⟩ KSP takes around 30 - 60% RAM with a browser open, and I have 12GB. If not accompanied with Mem Reduct it would have already crashed.
<queqiao-> ⟨Starcatcher⟩ +RSS/RO/RP1
<queqiao-> ⟨PrinceZola ùwú⟩ Really???
<queqiao-> ⟨PrinceZola ùwú⟩ I had no idea there were "memory usage optimizers"
<queqiao-> ⟨PrinceZola ùwú⟩ I got that alloc crash error many times already
<queqiao-> ⟨PrinceZola ùwú⟩ Let's see if things improve :o
<queqiao-> ⟨Marsh⟩ KSP uses 20+ for me, after a few hours of play it’s around 28
<queqiao-> ⟨Starcatcher⟩ I use Integrated Graphics, so not great experience.
<queqiao-> It has 2 times lower framerate in 4k graphics, and only have smooth gameplay in 2k, no rssve.
<queqiao-> ⟨egg⟩ ⟪PrinceZola ùwú⟫ but i wanted to try it out ⮪ If you want to try it out, why not try it out and see whether the performance impact is OK on your system, instead of trying to divine something from the conflicting reports of other users on very different systems?
<queqiao-> ⟨PrinceZola ùwú⟩ ⟪egg⟫ If you want to try it out, why not try […] ⮪ thats what im gonna do :P
<queqiao-> ⟨PrinceZola ùwú⟩ ⟪Starcatcher⟫ Seems like RSS hits me more on the RAM. […] ⮪ can you tell me more about those two "fixes"?
<queqiao-> ⟨PrinceZola ùwú⟩ ⟪egg⟫ If you want to try it out, why not try […] ⮪ is principia on ckan?
<queqiao-> ⟨PrinceZola ùwú⟩ bruh
<queqiao-> ⟨PrinceZola ùwú⟩ hello there
<queqiao-> ⟨PrinceZola ùwú⟩ this server is kinda of amazing
<queqiao-> ⟨Sumguy⟩ ⟪PrinceZola ùwú⟫ is principia on ckan? ⮪ No
<queqiao-> ⟨Stonesmile⟩ ⟪PrinceZola ùwú⟫ is principia on ckan? ⮪ Principia is not for CKAN, for multiple reasons; the main one is that Principia is written in C++ instead of C#
<queqiao-> ⟨Stonesmile⟩ * on
<queqiao-> ⟨面条参考系⟩ ⟪PrinceZola ùwú⟫ this server is kinda of amazing ⮪ Why yes, isn't it surprising that a server ran by devs has devs in it.
<queqiao-> ⟨Stonesmile⟩ There are other reasons like geographic region of users being visible with the current system. This revealed a large user base in China, which now has an easier way to download the mod
<queqiao-> ⟨PrinceZola ùwú⟩ whats the purpose of the master?
<queqiao-> ⟨面条参考系⟩ If you need to ask it's probably not for you.
<raptop> I'm pretty sure you're supposed to detach the master's head or something
<queqiao-> ⟨SpaceMarine⟩ its the alt history confederacy version
<queqiao-> ⟨SpaceMarine⟩ of the mod
<queqiao-> ⟨PrinceZola ùwú⟩ ⟪面条参考系⟫ If you need to ask it's probably not […] ⮪ if i need to ask maybe i havent read enough/wasnt clarified
<queqiao-> ⟨面条参考系⟩ Well, the readme does indeed tell you what you _should_ download and install.
<queqiao-> Serious answer: the master zip contains the source code for the mod should you want to build it yourself.
<queqiao-> ⟨PrinceZola ùwú⟩ Okey, thx for the response xP
_whitelogger has joined #principia
<queqiao-> ⟨egg⟩ ⟪面条参考系⟫ Why yes, isn't it surprising that a […] ⮪ to be honest this channel was here for a while before I joined.
<queqiao-> ⟨egg⟩ I found out about it in 2019, apparently.
UmbralRaptor has joined #principia
UmbralRaptor has quit [Client Quit]
UmbralRaptor has joined #principia
UmbralRaptop has quit [Ping timeout: 189 seconds]
UmbralRaptop has joined #principia
UmbralRaptor has quit [Ping timeout: 189 seconds]
<queqiao-> ⟨Starscraper⟩ ⟪PrinceZola ùwú⟫ Okey, thx for the response xP ⮪ Yeah, Github can be a bit confusing, it's geared more for devs than end-users. But the bottom of the main page has install instructions, iirc. Just follow those. But since you downloaded hermite I think you have been?
<queqiao-> Also good luck with 8gb. Be careful setting the number of calculation steps too high in the flight planner. I don't know much about how Principia is optimized but I'd assume that works harder on the RAM.
<queqiao-> ⟨PrinceZola ùwú⟩ the result was. okay i guess? but the game was running at 30 fps qwq
<queqiao-> ⟨PrinceZola ùwú⟩ at least it survived for a while
<queqiao-> ⟨PrinceZola ùwú⟩ so it was somewhat stable
<queqiao-> ⟨pEdro⟩ main impact from principia is fast timewarping and long load times for me
<queqiao-> ⟨Starscraper⟩ ⟪egg⟫ to be honest this channel was here for […] ⮪ Oh hey, the man himself.
<queqiao-> Like, do you know what the story behind the stock game's patched conics system is? Does the stock game do that because it's easier to implement, or for optimization?
<queqiao-> I was wondering some things about this awesome mod...
<queqiao-> Also I've worked with Unity quite a bit - so does this use compute shaders for the trajectories? But while the GPU is fantastic for parallel tasks, it doesn't seem like N-body trajectories are like, parallelizable at all.
<queqiao-> I understand the gist of numeric integration. I've passed computational physics courses where I've used the Runga-Kutta method, so ostensibly I understand it. I do understand Taylor expansions, though. You can take a Taylor expansion from a series of points by approximating the derivatives and perhaps that part is parallelizable but that doesn't give you the initial set of points.
<queqiao-> ⟨Transparent_Elemental⟩ I know it's possible to make parallel implementation of a program that solves parabolic PDE, so one would assume that it's possible to make that for other types of differential equations, if this is what you meant
<queqiao-> ⟨Starscraper⟩ Solves them analytically or numerically?
<queqiao-> I mean, there's no general solution to the n-body problem so it's gotta be numeric, unless it's doing an analytical approximation - which is what the Taylor series is, so I guess it _is_ doing analytical solving, somewhere.
<queqiao-> Huh. But given it solves the trajectory in steps, there's definitely a numerical integration component to it. I don't know how you can parallelize that, though, since the whole point of numerical integration is adding the points sequentially.
<queqiao-> ⟨Transparent_Elemental⟩ I don't understand differential equations (let alone parallel calculations) well enough to talk about analytical solution, but in my experience every time there's an analytical solution to anything it's always more expensive to compute it than implement numerical algorithm
<queqiao-> ⟨Transparent_Elemental⟩ also don't forget that taylor approximation is by definition not analytical because you can't sum an infinite number of terms of taylor expansion
<queqiao-> ⟨egg⟩ _stares_
<queqiao-> ⟨Transparent_Elemental⟩ also I think there has been some PRs to principia with like changes to matrix related algorithms, those could very well be paralleled, but whether principia does that or not I don't know
<queqiao-> ⟨Starscraper⟩ ⟪Transparent_Elemental⟫ also don't forget that taylor […] ⮪ Hah. Maybe I misunderstand the definition of "analytical", then.
<queqiao-> ⟨Starscraper⟩ Also, sorry to ask and run but I gotta hop in a meeting. I'll respond to things when I get back.
<queqiao-> ⟨Transparent_Elemental⟩ I mean it in a sense that it is an exact solution, there are a couple of different meanings to the word "analytical", yeah
<queqiao-> ⟨egg⟩ > Does the stock game do that because it's easier to implement, or for optimization?probably the former, I don’t think the developers would have known how to even begin to look at the numerical integration problem; but also it’s a design choice, it avoids a lot of complexity in gameplay as well.
<queqiao-> ⟨egg⟩ > confused speculation and speculative confusion about numerics… you both seem very confused.
<queqiao-> ⟨egg⟩ The GPU is not involved, except in drawing noodles on the screen.
<queqiao-> ⟨egg⟩ PDEs are _not_ like ODEs. PDEs are cursèd.
<queqiao-> ⟨egg⟩ Taylor expansions are irrelevant, except as a basic tool to prove things about the numerical methods involved.
<queqiao-> ⟨Transparent_Elemental⟩ ⟪egg⟫ > confused speculation and speculative […] ⮪ did I say something wrong? 🤔
<queqiao-> ⟨egg⟩ ⟪Transparent_Elemental⟫ I know it's possible to make parallel […] ⮪ you deduced something about methods for ODEs from something about methods from PDEs, which doesn’t reflect how these problems are treated.
<queqiao-> ⟨egg⟩ ⟪Transparent_Elemental⟫ I don't understand differential […] ⮪ Depends what you are doing, in some circumstances an analytic approach is better, we do that for rigid body rotation
<queqiao-> ⟨egg⟩ ⟪Transparent_Elemental⟫ also don't forget that taylor […] ⮪ correct, but also Taylor polynomials are not directly relevant to any reasonable way to approach the numerical integration problem
<queqiao-> ⟨egg⟩ ⟪Transparent_Elemental⟫ also I think there has been some PRs to […] ⮪ There are matrix algorithms, but they are not parallelized, nor are they involved in the numerical integration problem
<queqiao-> ⟨egg⟩ you deduced something about methods for ODEs from something about methods for PDEs, which doesn’t reflect how these problems are treated.
<queqiao-> ⟨Transparent_Elemental⟩ ⟪egg⟫ you deduced something about methods for […] ⮪ can't you make parallel calculations for either of these whenever you have, say, vector and matrix operations? it's not that the entire thing would be run in parallel, but some of the basic operations could be
<queqiao-> ⟨egg⟩ sure, there is some SIMD going on on the 3d stuff. No GPUs need be involved here, AVX is already wider than you need for that!
<queqiao-> ⟨egg⟩ That also has nothing to do with the reason why PDE solvers would tend to be parallel.
<queqiao-> ⟨Transparent_Elemental⟩ yeah, I wasn't trying to say that taylor approximation or matrix algorithms are somehow used or relevant since 1) you need to know the function to use taylor and you don't in PDEs and 2) the matrix algos were just an example
<queqiao-> ⟨egg⟩ > you need to know the function to use taylor and you don't in PDEsThe problem is not one of PDEs in any meaningful sense, but rather one of ODEs, and we do in fact know the derivative, that’s what you have the ODE (and we could express higher derivatives from that if we so wished).
<queqiao-> But Taylor expansions are just not how you deal with these problems.
<queqiao-> ⟨egg⟩ Starscraper was gesturing in the right direction by talking about Runge–Kutta methods.
<queqiao-> ⟨egg⟩ > you need to know the function to use taylor and you don't in PDEsThe problem is not one of PDEs in any meaningful sense, but rather one of ODEs, and we do in fact know the derivative, that’s what you have in the ODE (and we could express higher derivatives from that if we so wished).
<queqiao-> But Taylor expansions are just not how you deal with these problems.
<queqiao-> ⟨egg⟩ We have a bunch of integrators: https://github.com/mockingbirdnest/Principia/tree/master/integrators
<queqiao-> ⟨lamont⟩ N-body isn't very parallelizable unless you do the N log N stuff is it?
<queqiao-> ⟨egg⟩ Well, you could parallelize the force computations, but they are cheap individually and there aren’t _that_ many of them so it’s not worth it, and you then need to bring everything together everything every step. Or every stage, rather.
<queqiao-> ⟨lamont⟩ right
<queqiao-> ⟨lamont⟩ and if you're doing force computations on another core you need to move that memory around to the main core every step
<queqiao-> ⟨lamont⟩ probably faster to optimize for branch prediction and prefetching on a single core
<queqiao-> ⟨egg⟩ More useful is to parallelize the entire computations of the trajectories of multiple vessels (which, being massless, don’t interact). And we do that.
<queqiao-> ⟨Starscraper⟩ ⟪egg⟫ There are matrix algorithms, but they […] ⮪ Where I've heard of Taylor series being used in numerical integration isn't in the numerical integration part per sae, but I've heard of some professors at my univ. using Taylor series to approximate Relativistic gravity with equations that are much simpler (though have up to hundreds of auto-generated terms, they still have much simpler properties).
<queqiao-> ⟨Starscraper⟩ Theoretically, depending on how your code is compartmentalized, I wonder if it would be possible to add a few extra terms on the gravity equations to approximate relativity in KSP? 🤔 lol
<queqiao-> ⟨egg⟩ ah yes, limiting the _problem_ to lower orders is a thing. In fact we do that with the spherical harmonics expansion of the geopotential (going to higher order as you get closer).
<queqiao-> ⟨egg⟩ ⟪Starscraper⟫ Theoretically, depending on how your […] ⮪ We’ve looked into that for testing purposes (it would be entirely pointless in-game) but it’s a mess.
<queqiao-> ⟨egg⟩ Einstein–Infeld–Hoffmann was the thing we looked at.
<queqiao-> ⟨Starscraper⟩ In the Sol system, it would be pointless. But if some mod used some futuristic engines and took place around a black hole or such, it could be pretty cool. But I know that'd be a huge undertaking.
<queqiao-> ⟨egg⟩ If relativistic effects were not small, then low-order approximations wouldn’t be adequate. And then you start to have to deal with time properly, which is a right mess from a game design perspective.
<queqiao-> ⟨Starscraper⟩ Sadly, in multiplayer games we can't quite make time dilation work on the players behind the computers, lol.
<queqiao-> But "just" getting the trajectories right would be cool. If you _really_ wanted to be fancy, you could maybe even have a vessel-wide modifier on resource consumption and stress buildup and such to simulate time dilation.
<queqiao-> But I think even roughly approximate stuff would be pretty neat. Maybe just overlap linear approximations of Kerr solutions.
<queqiao-> ⟨Starscraper⟩ +(though KSP isn't _normally_ a multiplayer game, just a funny thought)
<paculino> Does Principia even work with multiplayer mods since they mess with time?
<queqiao-> ⟨Starscraper⟩ +(*overlap as in, just linearly adding them for each relativistic-scale body. Probably only a handful in the system, if not just 1.)
<queqiao-> ⟨Starscraper⟩ Hmm. Unfortunately it's probably not just a matter of changing gravity equations to get different topologies, otherwise a wormhole would be pretty cool - throw in a wormhole near Saturn to take vessels to a distant black hole system, heheh. Though teleporting vessels isn't impossible...
<queqiao-> ⟨Starscraper⟩ +I've gone pretty far afield, here, though.
<queqiao-> ⟨sichelgaita⟩ Either it's small effects, and they are invisible, and they don't make the game more interesting (think: Mercury perihelion). Or it's big effects (lensing, frame dragging, etc.) and they require fully solving Einstein's equations and reinventing the game.
<queqiao-> ⟨lamont⟩ > using Taylor series to approximate Relativistic gravity with equations that are much simpler
<queqiao-> ⟨lamont⟩ that's probably linearized GR
UmbralRaptor has joined #principia
<queqiao-> ⟨Soviet Onion⟩ to first order 😏
<queqiao-> ⟨Soviet Onion⟩ g = h + n
<queqiao-> ⟨lamont⟩ you can get nonclassical effects like gravitational radiation out of first order approximations
<queqiao-> ⟨Soviet Onion⟩ ⟪sichelgaita⟫ Either it's small effects, and they are […] ⮪ fully solving is probably not a good idea 😟
<queqiao-> ⟨egg⟩ ⟪sichelgaita⟫ Either it's small effects, and they are […] ⮪ and renaming the mod to _Feldgleichungen_
UmbralRaptop has quit [Ping timeout: 189 seconds]
<queqiao-> ⟨egg⟩ (and at that point you are in PDE land, and PDEs are nasty.)
<queqiao-> ⟨Soviet Onion⟩ ⟪egg⟫ (and at that point you are in PDE land, […] ⮪ _flashbacks to when my friend and I tried to solve 6 simultaneous PDEs by hand_ 😔
UmbralRaptop has joined #principia
<queqiao-> ⟨Soviet Onion⟩ and EFEs are 10 non linear simultaneous PDEs 😟
UmbralRaptor has quit [Ping timeout: 189 seconds]
<queqiao-> ⟨Soviet Onion⟩ probably will reduce after some symmetry and approximations
<queqiao-> ⟨Starscraper⟩ ⟪sichelgaita⟫ Either it's small effects, and they are […] ⮪ Only if you demand perfection or near-perfection.
<queqiao-> Linearlizing GR should make it not require anything more extensive than Principia does, and should be able to capture fun and interesting general relativistic effects.
<queqiao-> ⟨Starscraper⟩ _See that frame-dragging first-hand_
<queqiao-> ⟨egg⟩ > Linearlizing GR should make it not require anything more extensive than Principia does, and should be able to capture fun and interesting general relativistic effects.I strongly suspect that both parts of this sentence are false.
<queqiao-> ⟨Starscraper⟩ I mean, the point of linearizing GR is to, you know, _linearlize it_.
<raptop> The problem of course is that GR in a solar system conteggst requires precision considerations that are quite high. Also, reworking the rendering engine so you can travel to the sun's focus at 550 au and see cool things
<queqiao-> ⟨Soviet Onion⟩ ⟪Starscraper⟫ Only if you demand perfection or […] ⮪ I doubt solving the linearised field equations will be of similar effort as to what Principia does now
<queqiao-> ⟨egg⟩ (The former, because those equations cease to have the nice properties that we rely on, the latter, because, as sichelgaita wrote, because by the point these effects are significant, you are not capturing them adequately by a linearization.)
<queqiao-> ⟨Soviet Onion⟩ ⟪egg⟫ (The former, because those equations […] ⮪ ah, makes sense
<queqiao-> ⟨egg⟩ -because
<raptop> Let's see: the infamous 43"/century (and I guess modestly more interesting orbits for something like PSP, though what about photon pressure or plasma density?), GPS shenanigans (very high precision required), solar focus at 550 au. Any other relevant effects?
<queqiao-> ⟨Starscraper⟩ ⟪Soviet Onion⟫ I doubt solving the linearised field […] ⮪ Well... not "only as much effort" - but I don't think it'd take _overhauling the game's code_ any more than Principia does, I should say.
<queqiao-> ⟨Starscraper⟩ +deeply
<queqiao-> ⟨Soviet Onion⟩ ⟪raptop⟫ Let's see: the infamous 43"/century […] ⮪ photon pressure 😍
<queqiao-> finally, actually working photon sails
<raptop> Soviet Onion:Consider: the YORP (or at least Yarkovsky) effect
<queqiao-> ⟨Soviet Onion⟩ ⟪raptop⟫ Soviet Onion:Consider: the YORP (or at […] ⮪ 🥵
<queqiao-> ⟨Soviet Onion⟩ I wish RSS had some more asteroids and comets included 😔
<queqiao-> ⟨Starscraper⟩ ⟪egg⟫ (The former, because those equations […] ⮪ It depends on how closely "adequately" is. I'm thinking even just having the game take place on a planet orbiting a star orbiting a Kerr black hole - where the BH is relativistic but everything else is just Newtonian added to some relativistic approximations - would be good enough for a game.
<queqiao-> Now, it _would_ be super cool to have like, maybe a smaller 2-3-stellar mass BH and/or even a Neutron Star orbiting a supermassive black hole at the center of the system, with the player's star being relatively normal, and using the neutron star/black hole system for gravity assists, maybe. Maybe if you set it up right you could use linearlized approximations just fine (nothing too dense/massive too close to anything else too dense/massive).
<queqiao-> ⟨Starscraper⟩ -added to some relativistic approximations
<raptop> Ah, gravitational lensing angle is defined as θ = 4GM/rc². Converge distances look long compared with solar system size for main sequence stars (and lolgiants), though a small fraction of an au for WDs?
<raptop> Hrm, not sure how easy modeling the penrose process would be
<queqiao-> ⟨egg⟩ ⟪Starscraper⟫ It depends on how closely "adequately" […] ⮪ I think you are confusing multiple kinds of eliminations of higher-order effects. https://en.wikipedia.org/wiki/Linearized_gravity still puts you in PDE land, and at that point you are dealing with a problem that is incommensurably messier than anything Principia does. Conversely things like...
<queqiao-> ... https://en.wikipedia.org/wiki/Einstein%E2%80%93Infeld%E2%80%93Hoffmann_equations aren’t going to show you any fun effects, just make the perihelion precession of Mercury more accurate.
<raptop> So if you bring on ferram4 to help with the fun GR effects, the PDEs will be easier to deal with?
<queqiao-> ⟨lamont⟩ no