Re: [vos-d] Thought problem 2: physics 2
On 2/3/07, chris <[EMAIL PROTECTED]> wrote: > On 2/3/07, Reed Hedges <[EMAIL PROTECTED]> wrote: > > chris wrote: > > >> Of course... why not use a big integer for time? > > > > I would guess that lots of software does, especially since that's what > > most operating systems give you (e.g. time_t). > > > > > > > > A big integer at a fixed precision has larger relative error than a > > > small one > > > > Why? > because, although the resolution is even, the relative error = number/maxint. oops that's not right - I was thinking in terms of how you calc error for floating point :( I need coffee ... chris > > > > > > > Also, ODE doesn't use any randomness does it? > Well it might - I posted questions to the relevant forum but got not reply. > however, if I repeat the same experiment at, say the origin, the > results are repeatable. > > chris > > > > > > Reed > > > > ___ > > vos-d mailing list > > vos-d@interreality.org > > http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d > > > ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
Re: [vos-d] Thought problem 2: physics 2
On 2/3/07, Reed Hedges <[EMAIL PROTECTED]> wrote: > chris wrote: > >> Of course... why not use a big integer for time? > > I would guess that lots of software does, especially since that's what > most operating systems give you (e.g. time_t). > > > > > A big integer at a fixed precision has larger relative error than a > > small one > > Why? because, although the resolution is even, the relative error = number/maxint. > > > Also, ODE doesn't use any randomness does it? Well it might - I posted questions to the relevant forum but got not reply. however, if I repeat the same experiment at, say the origin, the results are repeatable. chris > > > Reed > > ___ > vos-d mailing list > vos-d@interreality.org > http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d > ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
Re: [vos-d] Thought problem 2: physics 2
chris wrote: >> Of course... why not use a big integer for time? I would guess that lots of software does, especially since that's what most operating systems give you (e.g. time_t). > > A big integer at a fixed precision has larger relative error than a > small one Why? Also, ODE doesn't use any randomness does it? Reed ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
Re: [vos-d] Thought problem 2: physics 2
On 2/2/07, Ken Taylor <[EMAIL PROTECTED]> wrote: chris wrote: > Thought problem 2: physics 2 > > Suppose I am going to do a rigid body simulation. I put one box (box1) > on a plane, at the origin and hold another box (box2) suspended a > meter above the plane nearby. I release box2 at time t=20 and it > bounces, perhaps collides with box1 then eventually comes to rest. I > snap an image of the rest state of the sim. > > Now I repeat the entire sim with viewpoint, boxes and plane in exactly the same > position as before. I drop box2 at time t=8000, let it bounce and snap an > image of the sim when it is at rest. > > Question: will the two images of the two experiments show box2 in the > same rest position relative to box1? > Well first I was going to say "of course not." But then I realized... time is probably floating-point as well, and has less precision when the number is bigger... oh snap! that's right - the results are different again - see attached. Of course... why not use a big integer for time? A big integer at a fixed precision has larger relative error than a small one so the physics calc will still produce different results. Depending on size of number and precision, an integer can be worse than a float, or vice versa. So, keeping an observation centered at the origin of both space and time can provide high fidelity *and* consistency. chris -Ken ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d dropBlock0cfTime.png Description: PNG image ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
Re: [vos-d] Thought problem 2: physics 2
I wrote: > chris wrote: > > Question: will the two images of the two experiments show box2 in the > > same rest position relative to box1? > > > > Well first I was going to say "of course not." I mean I was going to say "of course." I'm tired. -Ken ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
Re: [vos-d] Thought problem 2: physics 2
chris wrote: > Thought problem 2: physics 2 > > Suppose I am going to do a rigid body simulation. I put one box (box1) > on a plane, at the origin and hold another box (box2) suspended a > meter above the plane nearby. I release box2 at time t=20 and it > bounces, perhaps collides with box1 then eventually comes to rest. I > snap an image of the rest state of the sim. > > Now I repeat the entire sim with viewpoint, boxes and plane in exactly the same > position as before. I drop box2 at time t=8000, let it bounce and snap an > image of the sim when it is at rest. > > Question: will the two images of the two experiments show box2 in the > same rest position relative to box1? > Well first I was going to say "of course not." But then I realized... time is probably floating-point as well, and has less precision when the number is bigger... oh snap! Of course... why not use a big integer for time? -Ken ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
[vos-d] Thought problem 2: physics 2
Thought problem 2: physics 2 Suppose I am going to do a rigid body simulation. I put one box (box1) on a plane, at the origin and hold another box (box2) suspended a meter above the plane nearby. I release box2 at time t=20 and it bounces, perhaps collides with box1 then eventually comes to rest. I snap an image of the rest state of the sim. Now I repeat the entire sim with viewpoint, boxes and plane in exactly the same position as before. I drop box2 at time t=8000, let it bounce and snap an image of the sim when it is at rest. Question: will the two images of the two experiments show box2 in the same rest position relative to box1? ___ vos-d mailing list vos-d@interreality.org http://www.interreality.org/cgi-bin/mailman/listinfo/vos-d
