KSP2 EA Grand Discussion Thread.

[Periple]

9 minutes ago, MarcAbaddon said:

What do you mean by isolating it from the orbital mechanics? The orbital mechanics are already isolated from the current physics engine.

Unfortunately it's not -- I've had orbits move all over the place when rotating my craft with reaction wheels for example. That should have net zero impulse, but clearly because of physics inaccuracies it doesn't. K-drive effects. By isolating the systems I mean computing the engine impulse separately and ensuring that only that affects the trajectory. You don't need to rewrite or deeply rework the physics engine for that. They've already got the fundamentals for that – that's how they compute trajectories for unloaded craft (although they probably make simplifications that would need to be addressed for loaded craft, e.g. asymmetric thrust). 

(Solving this stuff in the physics engine would be just about impossible anyway, there's no way to have a physics engine precise enough that you could derive an interstellar-precision trajectory, and performant enough that you could use it for real-time rigid-body stuff.)

8 minutes ago, MarcAbaddon said:

It's really difficult to get everything working as you would like - there  is a reason why original KSP 1 or the modders never managed to fully tame the physics engine.

It's not trivial but it's certainly possible. KSP2 already has part of the solution with the partially-unloaded craft. They just need to get it all together, debugged, and working. Much harder on top of the KSP1 code of course!

Edited March 2, 2023 by Periple

[AlamoVampire]

@Periple um no. Multiplayer is far more complicated. Timewarp, parts counts, mod compatibility and so much more. Im done. <walks away>
140603022023

[Manul]

5 minutes ago, Periple said:

They don't need to rework the physics engine much, just isolate it from the orbital mechanics.

And rewrite everything to use doubles instead of floats (obviously it's impossible).  It's the range of a few kilometers where floating point calculations begin to fail, not hundreds or thousands km.   Most of Kraken attacks occur when a part with a small mass is attached to a part with a high mass, the most likely it's a floating point error causing near infinite forces that tear the craft apart.

[MarcAbaddon]

There's an autopilot namespace in the main .dll by the way.

Maybe you'll get lucky at some point (just taking a look myself). TBH I don't have found that many about colonies yet and the networking code seems more like generic things to setup initial communication. But my look so far was very superficial.

[MarcAbaddon]

10 minutes ago, Periple said:

Unfortunately it's not -- I've had orbits move all over the place when rotating my craft with reaction wheels for example. That should have net zero impulse, but clearly because of physics inaccuracies it doesn't. K-drive effects. By isolating the systems I mean computing the engine impulse separately and ensuring that only that affects the trajectory. You don't need to rewrite or deeply rework the physics engine for that. They've already got the fundamentals for that – that's how they compute trajectories for unloaded craft (although they probably make simplifications that would need to be addressed for loaded craft, e.g. asymmetric thrust). 

(Solving this stuff in the physics engine would be just about impossible anyway, there's no way to have a physics engine precise enough that you could derive an interstellar-precision trajectory, and performant enough that you could use it for real-time rigid-body stuff.)

It's not trivial but it's certainly possible. KSP2 already has part of the solution with the partially-unloaded craft. They just need to get it all together, debugged, and working. Much harder on top of the KSP1 code of course!

That's supposed to be a floating point bug related to changing reference local reference frames, and it can happen even if you don't rotate your craft. It's one of the most concerning bugs to me, but I don't think it's linked that tightly to the Unity physics, especially not RigidBody. Of course, there needs to be passover of information between the systems otherwise collisions couldn't change your orbit. 

[Periple]

23 minutes ago, AlamoVampire said:

@Periple um no. Multiplayer is far more complicated. Timewarp, parts counts, mod compatibility and so much more. Im done. <walks away>
140603022023

What!

[Vanamonde]

A number of overlapping threads have been merged. 

[Bej Kerman]

25 minutes ago, AlamoVampire said:

um no. Multiplayer is far more complicated. Timewarp, parts counts, mod compatibility and so much more

So it helps the game has been designed from the ground up to support it

[MarcAbaddon]

Looks like founding colonies will be done with a specific module (i.e. in the same sense that solar panels are modules of type solar panel). So there's probably going to be some kind of part you use for that.

[Manul]

1 hour ago, RocketRockington said:

Based on the fact they they only had a physics programmers at Intercept from 2021 to 2022, I think maybe even they didn't dmtry to do everything at once, they assumed they could copy KSP1 physics code without a dedicated programmers and focused on other things that we saw...for some reason they worked on the VAB and UI very early?  And the Kerbal animation?  Wish we knew what was really going on, in this case I want to know how the sausage was made.

Sounds scary. My most frustrating experience with KSP is when my precious aircraft carrier on Laythe or a huge base attached to the side of a Dres canyon suddenly explode for no reason on scene load and loading any saves doesn't help (the cheatmenu doesn't help either, with unbreakable joints I just get a dancing tornado of parts that never stops). Had to scrap many complicated designs that took me weeks to build and to get to some distant places, that was a huge waste of time (and funds because I do all this in career mode). My biggest hope for KSP 2 was that Kraken will be slain and such things will never happen again.

[AlamoVampire]

1 hour ago, Periple said:

What!

Timewarp. Parts counts, mod compatibility ALL make multiplayer more complicated. Please go through all the threads discussing how its far more complicated. But as my final pair of examples: your pc can handle a max part count of say 200 parts in physics range at once. Mine can comfortably handle say 1000 before it starts to even hint at strain. I can safely get say 60fps at my station that has with my current ship say 1250 parts. You get 60fps at a max of 200. You need fuel, i have it. Does my monstrosity make you go from frames per second to minutes per frame? Or how about time warp? Does my need to warp from LKO to an eeloo encounter and then to eeloo screw you up? Im warping YEARS into the future from your perspective. Are we no longer in the same temporal frame of reference? Did my multi year time war catastrophically screw you? This is but the tip of the complications. Ive said all this over and over and over year after year in thread after thread. Multiplayer is more complicated than adding mechjeb as mp has more things to consider. You dont see that? 
 

154203022023

154503022023

155103022023

Edited March 2, 2023 by AlamoVampire

[Jacke]

One thing that I'm kind of sad wasn't including in KSP2 was an expanded Kerbol system.  It just has the same bodies as stock KSP1.  Having more like in the Outer Planets Mod would be just great.  This system should provide more before even going interstellar.

[Bej Kerman]

13 minutes ago, AlamoVampire said:

Timewarp. Parts counts, mod compatibility ALL make multiplayer more complicated. Please go through all the threads discussing how its far more complicated. But as my final pair of examples: your pc can handle a max part count of say 200 parts in physics range at once. Mine can comfortably handle say 1000 before it starts to even hint at strain. I can safely get say 60fps at my station that has with my current ship say 1250 parts. You get 60fps at a max of 200. You need fuel, i have it. Does my monstrosity make you go from frames per second to minutes per frame? Or how about time warp? Does my need to warp from LKO to an eeloo encounter and then to eeloo screw you up? Im warping YEARS into the future from your perspective. Are we no longer in the same temporal frame of reference? Did my multi year time war catastrophically screw you? This is but the tip of the complications. Ive said all this over and over and over year after year in thread after thread. Multiplayer is more complicated than adding mechjeb as mp has more things to consider. You dont see that? 

Those are all problems that will either be eliminated or eased by how KSP 2 has been built

[MarcAbaddon]

1 minute ago, Bej Kerman said:

Those are all problems that will either be eliminated or eased by how KSP 2 has been built

I don't see how we would know that at this point. 

[RocketRockington]

14 minutes ago, AlamoVampire said:

Timewarp. Parts counts, mod compatibility ALL make multiplayer more complicated. Please go through all the threads discussing how its far more complicated. But as my final pair of examples: your pc can handle a max part count of say 200 parts in physics range at once. Mine can comfortably handle say 1000 before it starts to even hint at strain. I can safely get say 60fps at my station that has with my current ship say 1250 parts. You get 60fps at a max of 200. You need fuel, i have it. Does my monstrosity make you go from frames per second to minutes per frame? Or how about time warp? Does my need to warp from LKO to an eeloo encounter and then to eeloo screw you up? Im warping YEARS into the future from your perspective. Are we no longer in the same temporal frame of reference? Did my multi year time war catastrophically screw you? This is but the tip of the complications. Ive said all this over and over and over year after year in thread after thread. Multiplayer is more complicated than adding mechjeb as mp has more things to consider. You dont see that? 

All that is true - but they've been promising multiplayer since the inception of the project.  You'd think at least on the timewarp end, they've got some workable solution to it, or it wouldn't have been something they mentioned so early.

The fact that the release version is missing, according to people who inspected the code, network synch for objects, does not say 'robust multiplayer' to me.  Maybe multiplayer isn't going to be what we think it is?  Maybe multiplayer will be, I dunno, just resource sharing at colonies and ships never come in contact with each other.  Something that requires only very limited synchronization.

2 hours ago, RayneCloud said:

 

Hmmm.. if you say so.

Oh that's nice - I guess they've started support for Flex again.  Flex seems to get off-again, on-again support from PhysX, last I knew it was 'off again' but I guess it's on again now.  Though that video is 6 years old, you should have linked something more recent like this - https://developer.nvidia.com/flex-example

[Alexoff]

[HebaruSan]

34 minutes ago, AlamoVampire said:

Are we no longer in the same temporal frame of reference?

At this point, if multiplayer does get implemented, that's exactly what I'm expecting: Players will share real-world time only, with simulation fidelity left out in the cold. Other players' vessels will speed up and slow down as they enter and exit time warp.

[RayneCloud]

9 minutes ago, HebaruSan said:

At this point, if multiplayer does get implemented, that's exactly what I'm expecting: Players will share real-world time only, with simulation fidelity left out in the cold. Other players' vessels will speed up and slow down as they enter and exit time warp.

Wildly speculating, with very little information., is not something I find helpful or productive.

Edited March 2, 2023 by RayneCloud

[Bej Kerman]

21 minutes ago, Alexoff said:

I run on similar enough hardware and get quite a bit more frames than this sheet would indicate.

[RayneCloud]

Just now, Bej Kerman said:

I run on similar enough hardware and get quite a bit more frames than this sheet would indicate.

shhhh... you're not pushing the right narrative... lol

[HebaruSan]

5 minutes ago, RayneCloud said:

Wildly speculating, with very little information., is not something I find helpful or productive.

It can be fun, though, and it's a novel way of summarizing our impressions.

After thinking about it a bit more, that kind of simple implementation might not even be so bad. You could still meet up with your friends at a space station for EVA hijinks, but everybody could time warp however they liked. Include some kind of "blur" visual effect for ships moving in a different timeflow than you, and it might even be kind of neat.

[RayneCloud]

5 minutes ago, HebaruSan said:

It can be fun, though, and it's a novel way of summarizing our impressions.

After thinking about it a bit more, that kind of simple implementation might not even be so bad. You could still meet up with your friends at a space station for EVA hijinks, but everybody could time warp however they liked. Include some kind of "blur" visual effect for ships moving in a different timeflow than you, and it might even be kind of neat.

We'll have to wait and see. I'm hopeful for something at least usable and neat considering they've mentioned in interviews one person being in one star system and one being in another and them sending each other stuff via delivery routes. 

[cocoscacao]

KSP 2 devs and EA players

[The Aziz]

1 hour ago, Alexoff said:

Can you do a comparison with KSP 0.7.3?

Who would've thought that 10 year old game runs faster, what a surprise.

[Manul]

4 minutes ago, The Aziz said:

Who would've thought that 10 year old game runs faster, what a surprise.

Unoptimized 10-year old game with lots of mods. With all these mods it wouldn't even load on my 8 year old PC that was retired because it couldn't run KSP any more.