Orionsarm.org

[kojikun]

It's a standard for measuring the activities of a planetary body; that's not unsubstantiated at all. There's nothing arbitrary about it; my earlier estimates which I was going to throw out(1GFLOP for a planet) would be pulled out of the air.

You need an entire supercomputer to model a single planets gravitational effect? I don't think so.

If you don't like it, Kojikun, provide reasoning to throw it out.

You're the one who needs to provide some reasoning for USING it.

[SirNitram]

It's a standard for measuring the activities of a planetary body; that's not unsubstantiated at all. There's nothing arbitrary about it; my earlier estimates which I was going to throw out(1GFLOP for a planet) would be pulled out of the air.

You need an entire supercomputer to model a single planets gravitational effect? I don't think so.

And it's movement. Again, if you have something more reasonable, show it.

If you don't like it, Kojikun, provide reasoning to throw it out.

You're the one who needs to provide some reasoning for USING it.

I've given that, dickhead. Unlike you in this thread, I've backed up what I've presented. Just because it's setting fire to your favorite sacred fucking cow doesn't really bother me. Are you going to provide any serious rebuttal?

[kojikun]

I've given that, dickhead. Unlike you in this thread, I've backed up what I've presented. Just because it's setting fire to your favorite sacred fucking cow doesn't really bother me. Are you going to provide any serious rebuttal?

Where is your reasoning for a supercomputer per object? Nowhere. I'm not debating this, I'm simply saying that you've not given any reason why R2 would need that much computing power per object. Motion is simple, gravity is simple. You don't need an entire supercomputer for a single frelling planet or star.

[Hotfoot]

In order to prevent TK dickheads from messing with us, Mike made it so that any person registering using a free email account must make a donation as a sign of sincerity.

Okay...but the email address he has listed does not appear to be a free account, since ihug.au is not a free email service like hotmail or yahoo mail. Also, he has his own website...what is stopping him from using @orionsarm.com or somesuch?

[kojikun]

Don't know, but thats how it seems to be, Hotfoot.

[SirNitram]

I've given that, dickhead. Unlike you in this thread, I've backed up what I've presented. Just because it's setting fire to your favorite sacred fucking cow doesn't really bother me. Are you going to provide any serious rebuttal?

Where is your reasoning for a supercomputer per object? Nowhere. I'm not debating this, I'm simply saying that you've not given any reason why R2 would need that much computing power per object. Motion is simple, gravity is simple. You don't need an entire supercomputer for a single frelling planet or star.

Here's the reasoning: It's what I found being used to simulate Earth as accurately as possibel in reality. If you refuse to accept it, prove it too much. Christ, debating with your is like pulling teeth.

[kojikun]

Here's the reasoning: It's what I found being used to simulate Earth as accurately as possibel in reality. If you refuse to accept it, prove it too much. Christ, debating with your is like pulling teeth.

EarthSimulator simulates weather systems, dumbass. And you're making the claim that it's necessary, burden of proof is on you.

[Mad]

A quick Google search shows that there's at least 200 billion stars in the milky way. The SW galaxy is larger, but we'll just use ours (unless someone has a good source on the number of stars in the SW galaxy).

So we need to store the positions, velocities, and mass of 2E11 stars. Storing position, velocity, and mass would then require 7 numbers. That's 1.4E12 elements of information, for stars alone (not including planets, asteroid fields, and other potential hyperspace hazards). These figures are going to be huge, and will need to be precise, so given the scale we're talking about, a 64 bit number may not be precise enough. But we'll just assume they're using 64 bit numbers anyway. A 64 bit number takes up 8 bytes. So that'd require, at a minimum, 1E13 bytes, or 10 terabytes. Okay, fine, whatever; we know SW computers can handle that since they can store yottabytes of info on Coruscant.

But what about calculating the positions of the stars?

They need to calculate gravitational influences. That's going to be O(n^2), since every star is affected by every other star's mass, and, collectively, that adds up. That's the positions and the masses, for (4 * 2E11)^2, or 6.4E23 calculations. The computer then has to predict the star positions throughout the entire hyperspace jump. The course through hyperspace isn't going to be linear, either. We're looking at zetahertz processing low-end for an astromech droid/navicomputer (assuming 5 minutes to calculate gravitational influences and nothing else).

It should be noted that there are a number of possible optimizations I ignored since I don't know how accurate the calculations have to be. I also ignored a lot of other factors that will have to be computed, so hopefully it evens out. I haven't heard of a droid going quiet during calculations, either, which should also pump up the numbers a bit.

[kojikun]

Mad, you seem competent, so I have two questions: How many EarthSimulators would you guess that to be, and why couldn't SirNitram provide this kind of explanation?

[Mad]

Mad, you seem competent, so I have two questions: How many EarthSimulators would you guess that to be, and why couldn't SirNitram provide this kind of explanation?

I don't know anything about an EarthSimulator, so I dunno. And if SirNitram isn't a CompSci student, then I wouldn't expect him to think in terms of computational complexity.

(And, as a CompSci student, I should be thinking of how to optimize that nasty O(n^2) into something a bit nicer, but I don't know enough specifics about what is going on to know what kinds of optimizations are possible.)

[SirNitram]

Mad, you seem competent, so I have two questions: How many EarthSimulators would you guess that to be, and why couldn't SirNitram provide this kind of explanation?

I don't know anything about an EarthSimulator, so I dunno. And if SirNitram isn't a CompSci student, then I wouldn't expect him to think in terms of computational complexity.

Pretty much. And I dared to think Kojikun referencing a computer system in a debate about the positions of planets might denote a system about the positions of planets. Silly me.

[kojikun]

Pretty much. And I dared to think Kojikun referencing a computer system in a debate about the positions of planets might denote a system about the positions of planets. Silly me.

I was renferencing the EarthSimulator as a starting point for future supercomputers. You were the one who started using it for your little "R2 is s00p3r!!!11" arguement, and after repeated requests for an explanation why, the best I got back was "decause thats what I determined to be necessary, end of story." It's not my fault you're being a cunt.

[SirNitram]

Pretty much. And I dared to think Kojikun referencing a computer system in a debate about the positions of planets might denote a system about the positions of planets. Silly me.

I was renferencing the EarthSimulator as a starting point for future supercomputers. You were the one who started using it for your little "R2 is s00p3r!!!11" arguement, and after repeated requests for an explanation why, the best I got back was "decause thats what I determined to be necessary, end of story." It's not my fault you're being a cunt.

Yea, I'm such a cunt because I demand you back your shit up. What a horrible man I am.

[kojikun]

Yea, I'm such a cunt because I demand you back your shit up. What a horrible man I am.

Back what up?! I wasn't claiming that you DON'T need a supercomputer, I asked you why you said you DO. Fucking read, dumbass. Would you like me to quote my post requesting this?

Who says you need to have EarthSim power for EACH object? What reasoning is there for this?

Notice how I just asked why? Thats called NOT debating the point, just asking for a reason. You're the one saying that _I_ need to proove its NOT necessary. I was just asking why. Thats why you're a cunt, because you ARE.

[SirNitram]

A quick Google search shows that there's at least 200 billion stars in the milky way. The SW galaxy is larger, but we'll just use ours (unless someone has a good source on the number of stars in the SW galaxy).

300 billion(Roughly..) based on the Episode II ICS. The math for this is a few posts back, but it really changes very little.

So we need to store the positions, velocities, and mass of 2E11 stars. Storing position, velocity, and mass would then require 7 numbers. That's 1.4E12 elements of information, for stars alone (not including planets, asteroid fields, and other potential hyperspace hazards). These figures are going to be huge, and will need to be precise, so given the scale we're talking about, a 64 bit number may not be precise enough. But we'll just assume they're using 64 bit numbers anyway. A 64 bit number takes up 8 bytes. So that'd require, at a minimum, 1E13 bytes, or 10 terabytes. Okay, fine, whatever; we know SW computers can handle that since they can store yottabytes of info on Coruscant.

Works, though I'd throw in consideration that each system would have between hundreds and tens of thousands of objects that need to be considered. That'll push the whole thing up orders of magnitude, though this is fine for conservative things.

One consideration: Wouldn't you need 3 64 bit numbers per object? For it's position in a XYZ grid?

But what about calculating the positions of the stars?

They need to calculate gravitational influences. That's going to be O(n^2), since every star is affected by every other star's mass, and, collectively, that adds up. That's the positions and the masses, for (4 * 2E11)^2, or 6.4E23 calculations. The computer then has to predict the star positions throughout the entire hyperspace jump. The course through hyperspace isn't going to be linear, either. We're looking at zetahertz processing low-end for an astromech droid/navicomputer (assuming 5 minutes to calculate gravitational influences and nothing else).

I can't really comment on the accuracy here; Big O notation to me is something to do with a cartoon. But yea, one would have to know the positions throughout the jump, and the 5 minute thing is broadly correct. If there's more, well, we don't know.

It should be noted that there are a number of possible optimizations I ignored since I don't know how accurate the calculations have to be. I also ignored a lot of other factors that will have to be computed, so hopefully it evens out. I haven't heard of a droid going quiet during calculations, either, which should also pump up the numbers a bit.

Looks far more accurate than my blind guess.

[SirNitram]

Yea, I'm such a cunt because I demand you back your shit up. What a horrible man I am.

Back what up?! I wasn't claiming that you DON'T need a supercomputer, I asked you why you said you DO. Fucking read, dumbass. Would you like me to quote my post requesting this?

Who says you need to have EarthSim power for EACH object? What reasoning is there for this?

Notice how I just asked why? Thats called NOT debating the point, just asking for a reason. You're the one saying that _I_ need to proove its NOT necessary. I was just asking why. Thats why you're a cunt, because you ARE.

I gave my reason. Because you refused to point out what EarthSim was after bringing it up isn't my fault; yea, I'm a fucking horrible cunt for hoping you were on topic. Fine; I was mistaken. Amusingly, Mad's estimation seems to be even higher than mine.

[kojikun]

Looks far more accurate than my blind guess.

ahh well atleast you now admit that you pulled the numbers out of your ASS.

[SirNitram]

Looks far more accurate than my blind guess.

ahh well atleast you now admit that you pulled the numbers out of your ASS.

I suppose you think you're clever. The very last number in the calculation was based off you being misleading; woop de shit! The rest is still applicable.

[phongn]

The Earth Simulator is a 35 TFLOP supercomputer, by way of example. IBM's Blue Gene/L is expected to hit around 360 TFLOP in 2004-2005; Blue Gene will be ~1 PFLOP upon completion.

[kojikun]

I suppose you think you're clever. The very last number in the calculation was based off you being misleading; woop de shit! The rest is still applicable.

Misleading? How is this misleading?:

The earth simulators NEC processors run at 500MHz and get 8GFLOPS, so thats ~ 16FLOPS/Hz/processor.

I was getting a FLOPS measurement. Its not my fault you saw it and took it to be my saying it was a gravity and motion predictor. Thats your fault, not mine. That was in no way misleading. You're just a git.

[SirNitram]

Updated number crunching. The first part is unchanged, as it's accurate.

From the Ep II ICS, there are 300,000,000 stars in the Naboo sector. Ignoring all canon info about Naboo being a minor sector, we'll assume this is average.(3e8)

There are 1,000 sectors in Known Space.(3e11)

Using the real solar system as a guide, it seems safe to say there's at least a hundred objects worth keeping an eye on; depending on the tolerances of hyperdrive, it may skyrocket above that.(3e13)

O(n^2) should, unless I miss what O stands for, be (4 * 3e13)^2, or 1.2E27 calculations. Don't know how to work out Hz from here, Mad, some help?

[kojikun]

Keep up with this reasoning and you'll actually have an answer for my question. Provide that and I might give you a kiss. Cause I'm mean.

When they said minor, could they not have meant politically insignificant? Like, for instance, Brazil, Canada, India.. they're all minor countries, nothing as major as, say, the USA, Britain, or China, but they ARE large.

[SirNitram]

Keep up with this reasoning and you'll actually have an answer for my question. Provide that and I might give you a kiss. Cause I'm mean.

I'd settle for something substantiated in OA's favor; debates are so boring when only one side will do the math.

When they said minor, could they not have meant politically insignificant? Like, for instance, Brazil, Canada, India.. they're all minor countries, nothing as major as, say, the USA, Britain, or China, but they ARE large.

A potential, which is why I'm still listing it as average. Ultimately, the difference isn't that great; it's the orders of magnitude where things will tilt in one side or another.

[Mad]

300 billion(Roughly..) based on the Episode II ICS. The math for this is a few posts back, but it really changes very little.

It'd change things, but not drastically. I can try to keep it in mind if I need to redo the calcs, though.

Works, though I'd throw in consideration that each system would have between hundreds and tens of thousands of objects that need to be considered. That'll push the whole thing up orders of magnitude, though this is fine for conservative things.

Now that I think about it, a very probable optimization would be that each star system is considered as a whole when doing the gravitational calcs. Hazards within a star system would only need to be considered when a ship gets near one.

One consideration: Wouldn't you need 3 64 bit numbers per object? For it's position in a XYZ grid?

Already considered. The 7 numbers I said that were required are: x, y, and z coords for position, x, y and z vectors for velocity, and the mass of the system.

I can't really comment on the accuracy here; Big O notation to me is something to do with a cartoon. But yea, one would have to know the positions throughout the jump, and the 5 minute thing is broadly correct. If there's more, well, we don't know.

Big O notation isn't too bad, though deriving one from some arbitrary algorithm can be a pain. Just imagine how the number given would be graphed if it were a function and you'll get an idea as to how many more CPU cycles are required as the data increases. (x axis is amount of data, y axis is CPU cycles.)

As for proving that it's O(n^2)... take 4 objects and compare their gravitational influences... It'll take you 16-4=12 calculations. Take 8 and do the same thing. 64-8=56 calcs. So 300 billion stars would require (300 billion)^2 - 300 billion calcs... n^2 - n estimates to n^2.

If you can imagine ways to cut down on the calculations where the resulting inaccuracies won't affect the end results, then you might be able to reduce the workload a bit, I dunno how much as of yet, though. And, like I said, we're neglecting a lot of other calcs, and we don't know how often the graviational influences have to be recalculated during the processing, if at all.

[SirNitram]

On optimization:

In theory, you could get the shortcut of turning an entire sector or so into one example(Say, if you're moving entirely on the far side of the Galaxy), though of course these shortcuts kick your margin of error up each time. The 100 object thing was assuming we can lump together large chunks of asteroid belts as one particularly spread-out mass.