[fortytwo]

Re-read the argument that's being made. I already pointed out several times that the complexity of collision detection would remain the same (in each case, you'd have to consider the distance between your object and the Earth, paying attention to their shapes, rotation, etc.) - it's the complexity of motion where enormous savings can be made - both in terms of time and space.

If you move a complex object, every point of this object is in a different position after the movement, so every point needs to be updated.
This can be either done by ...
... either calculate each point separately (unusual but maybe more efficient for FE simulation because for each time slice it is the same value added to the z axis for each point).
... or calculate the center and orientation of the object and derive other points from it.
The only difference is where the movement for each point is handled - either by the moving function or by the collision detection function.
It's just a shift of the effort from one place to the other but it does not reduce it.

No, you just hyperfocused on replying without processing what you're replying to. You're also the victim of an oversimplified example, but I don't want to hammer JSS for making it simple too much. Once you consider things like parallelisation, the efficiency rift becomes even larger.

That's a (quite common) misconception of parallelism. Parallelism can make things faster, if it is well implemented. But parallelism makes an algorithm never more efficient (which would be less computing power for same result), it adds more overhead complexity.
This might change with quantum computers - but this will also change the algorithm in a way which makes this discussion obsolete.

Neither of those are necessary - all you need to do is periodically reset your co-ordinate system.

Which is the less smart version of "permanently adjust it to the center".

However, given everything we know about time complexity, space complexity, and our state-of-the-art algorithms, RET's case for being more efficient is non-existent.

I would not even dare to guess which one is more complex.

[Pete Svarrior]

The only difference is where the movement for each point is handled - either by the moving function or by the collision detection function.
It's just a shift of the effort from one place to the other but it does not reduce it.

I can see why you'd intuitively perceive it that way, but it's incorrect - handling it in motion needlessly duplicates near-identical calculations, since collision detection will have to transpose those points into relative co-ordinates for the second time.

That's a (quite common) misconception of parallelism. Parallelism can make things faster, if it is well implemented. But parallelism makes an algorithm never more efficient (which would be less computing power for same result), it adds more overhead complexity.

Once again, you are missing the point of what I said. I did not say that parallelisation would make an algorithm more efficient (I was talking to a person who clearly can write some code - it was my hope I wouldn't need to waste my time detailing the basics). I said that the complexity rift between a parallelised version of the RE model and a parallelised version of the FE model would be even more egregious than that between their single-threaded equivalents.

Which is the less smart version of "permanently adjust it to the center".

If you consider optimisation "not smart", then we can't have a meaningful argument about optimisation, as I'm sure you'll agree. Redundantly repeating an operation every tick is less efficient than only executing it when needed (in general - I'm sure you could find exceptions).

I would not even dare to guess which one is more complex.

Why are you arguing a position you would not even dare to guess, especially with such fervour? Seems counter-productive.

[fortytwo]

I can see why you'd intuitively perceive it that way, but it's incorrect - handling it in motion needlessly duplicates near-identical calculations, since collision detection will have to transpose those points into relative co-ordinates for the second time.

The point is that you do not need to move and/or transpose the points if you keep the earth object static relative to the vector space. So it's not about having one or two transformations (which is question of optimization) but zero. The "points" of the earth object need only to be updated when the shape (terrain) changes, otherwise they are fixed within the coordinate system.
Same is true for all other objects which are standing stable on earth, they can be treated as static until any other force than UA impacts them.

However, as you already pointed out, this approach of simulation (collision of meshes) would be far to simple for a simulated earth as we observe it - regardless of its shape.

Why are you arguing a position you would not even dare to guess, especially with such fervour? Seems counter-productive.

I stepped into the argumentation whether it is more efficient to keep earth static relative to the vector space or not - and I'm convinced static would be less complex (based on current techniques).

I think we don't have sufficient knowledge to judge the complexity of a earth simulation, regardless if it is flat or round, so I don't even try to compare FE vs RE in this context.

[Regicide]

I agree with forty-two: any simulation on the scale of the one being discussed is just completely beyond our current abilities, and it's likely meaningless for us to judge it based on current methods of computation. I'd also likely want to point out that if a concerned party was interested in creating an optimized scenario, they would probably look to other things for optimization, like giving us less cells, or making hairs larger. There are quite a few things that could be optimized with a much smaller effect.

[Pete Svarrior]

The point is that you do not need to move and/or transpose the points if you keep the earth object static relative to the vector space. So it's not about having one or two transformations (which is question of optimization) but zero. The "points" of the earth object need only to be updated when the shape (terrain) changes, otherwise they are fixed within the coordinate system.

You just moved your inefficiency around - you didn't address it at all.

I stepped into the argumentation whether it is more efficient to keep earth static relative to the vector space or not - and I'm convinced static would be less complex (based on current techniques).

Well, I can do nothing more than explain why that's incorrect. I guess we'll just have to agree to disagree, and the field of computing is going to continue not to care about our disagreement.

[JSS]

I agree with forty-two: any simulation on the scale of the one being discussed is just completely beyond our current abilities, and it's likely meaningless for us to judge it based on current methods of computation. I'd also likely want to point out that if a concerned party was interested in creating an optimized scenario, they would probably look to other things for optimization, like giving us less cells, or making hairs larger. There are quite a few things that could be optimized with a much smaller effect.

The thing about simulations is you don't have to simulate it all.  It doesn't actually matter where all your cells are until a doctor takes a sample. The program can just abstract it all away, and when you get a blood sample and it gets put into a machine, it can just look up your 'stats' and throw the right numbers onto the display.  Details don't need to be calculated until we look at them. The computer can just make things up as needed.

On a deeper level, you don't even need to 'render' much of the universe, even when people are looking right at it.  Just inject it right into our consciousness. The structures in out brains that remember things are all very self referential, you don't need to draw a complicated object, just put the idea of that object into a persons head.

If I dump a bucket of marbles down a deep well with my eyes closed, the computer wouldn't even need to simulate that.  Some random marble noises is all it needs.  If I walked away and the well collapsed a year later and nobody ever dug it up, the computer just has to store that a bunch of marbles were dropped in, just in case someone does eventually dig there.  Even then, if it's 200 years later and no records exist of those marbles it can just delete them entirely. Who would notice?

[JSS]

I think we don't have sufficient knowledge to judge the complexity of a earth simulation, regardless if it is flat or round, so I don't even try to compare FE vs RE in this context.

Considering the fact that the external universe doesn't have to follow our laws, it's completely pointless to even try and guess what their capabilities are.

If the entire universe is simulated then we have zero data to go on, it's all guessing and making stuff up.

Fun, but ultimately pointless.