Why on earth would you write assembly code unless you need it to speed up calculations?
There are no forces. Yes, “speed vs Density”, but then also angular trajectory vs gradient.
But one step at a time.
The Ab is the point-by-point ambient density as a result of an amassed affectance. So the speed of each afflate is determined by it relative position to that mass and the combination of its own density and that of the ambient at each point it traverses.
Bad choice of words on my part regarding artifact. You are correct in that a specific pattern of motion is dependent on a specific density.
Of course the density has to evolve into that specific density configuration first.
That is fair James.
Back to the field emulation…
Again, this is a time when you have some freedom to pick and choose which way you would like to setup the field. You have specific afflate characteristics defined as if they were marbles (not yet “fuzzy”), yet the field has no such hard objects within. So if asked what the field density is as some point, {d,e,f}, precisely how are you going to use the local afflates to determine (to declare) the precise density … at any point chosen?
I appreciate the freedom.
Point? That would require some sort of averaging. The word point is throwing me a little.
There are many legit reasons. But yes to avoid the overhead of abstract functions and bad optimization on the part of the compiler writer. Binary operations like adding and subtracting I imagine would be good for building massive fields of affectance.
By “point”, I merely mean a chosen location within the space. And yes, choose a type of averaging that will yield the kind of field you are trying to form (smoother, choppier,…).
And you will want to make that a “hook” so that you can play with it later.
A point can be chosen with any resolution I guess, obviously there is more than two “things” involved but points are infinitely small so I could base my average on a small or large selection.
Seems like the least of your worries.
I doubt it. You’re going to have a lot of calculations involving 1/x, 1/x^2 or 1/x^3 which will take up more of the processing.
I like the way you put that all of that.
You could, just for example, have a linear decrease in afflate characteristic with distance from surrounding afflates. Or it could be exponential, giving a more choppy effect. I would recommend a very limited local distance for the range of evaluation so as to not consume too much processor time.
I had to think for a while to figure a way to quickly evaluate each afflate’s (or location point’s) immediate surrounding. Of course, you are free to choose your own method.
Why do you say that?
You are not looking at it the way I am.
You are going to be needing a lot of:
$$dx = Af1.x - x$$ $$dy = Af1.y - y$$ $$dz = Af1.z - z$$
$$d = \sqrt{dx^2 + dy^2 + dz^2}$$
Cause compilers are pretty good and flexible these days. Even the old school coders are abandoning assembler because they don’t get a significant payback any more.
Your big problem is figuring out the fuzzball interactions.
Yeah, obviously. I personally would get a handle on the interaction of a couple (or handful) of fuzzballs before I worried about 200,000 fuzzballs.
Fair enough.
I do have an idea but for the life of me I can not seem to get it into words at the moment - my head is in the coding. I will code some interactions up then explain them. I have to go with my gut I think - if I get into trouble perhaps you can point out why.
Plus you put me off phyllo 
just messin with ya.
What does this notation mean?
Realize that the afflates do not interact with each other directly (phyllo’s concern). Each afflate interacts with it’s immediate ambient field, which is an indirect association with the combination of the other local afflates. Each afflate interacts with the averaged affect of all local afflates.
- dx = difference in x values, “differential”.
- Af1.x = x coordinate of Afflate1 (then Af21.x, Af53.x,…)
- x = x coordinate of the point of interest.
- d = distance between point of interest and a local afflate.
Actually, I guess that your notation will be more like, “Af.1.x” or some such.
One could do the simple yet stupid thing and just search the entire list of afflates to see which ones are close enough to be concerned with. To save processor search time, I chose to divide the entire cubic space into many cubical regions, 40x40x40 regional cubes. Every time an afflate moved, I recorded which region it was in. Then each region had a list of the only afflates of concern. Then in order to save memory space (perhaps not relevant in your case), I created a set of dynamic link-list functions to prevent wasted space (AddToLinkList, DeleteFromLinkList,…). Each afflate remembers the region. Each region stores an entry point into the link-list for all afflates. The search time turned out to be very quick without wasting memory resource.
Of course going down the link-list, each associated afflate had to be measured for closeness and varied trial formulas were used to decide how much affect each close afflate would have on the average at the point of interest (the afflate under calculation). The size of the afflates is very relevant at this point.
An additional issue is raised when you realize that one cannot merely examine a single region in order to average in all near afflates. Depending on how close the afflate is to a regional border, adjacent regions must also be included in the averaging. Unfortunately that requires the process to be considerably longer. I used a special table to provide quick region adjacency pointers, but the processor still had to go through each adjacent region’s link-list.
If I remember right, other than the video encoding, that examination of the local region to resolve the ambient affectance took most of the processor time. Once the ambience is resolved (still involving much more than currently discussed), how to move to afflate is very complex, involving the more sophisticated math (trig functions and such), but not terribly time consuming.