Say the details aren’t important, so we use the following logic,
at t1, a is 0, locus is 0
at t2, a is F/m, locus is v.t2
etc.
where they intersect, hit or miss, is v.tN’ and a.tN.tN’. where location of both colliding objects is n1 and n2.
The computer, would take t, n, and v, a as symbols. The processing is simple as t is a running total, n is a fv, and a is v.t.t’ where v is constant.
Such a machine need only consider t++, and the multiplication of v and t in the same clock.
But this gets interesting when we are faced with multiple projectiles as is common nowadays in games. This includes the modeling of a sprite running and firing bullets.
We can see that we need only implement fv over the whole game scene. Same with weather forecasts, but with a few more variables, namely, Work = energy.t , and it is incorporated as such that we have the following variables, force/t = mass.t while F is also equal to ma. We need only add F/t and m.t into our processing. These are simply vectors which alter fv’s vectors.
Say there is reflection, mirror effect, then we need to take fv and complicate it with n/n’. So far so good, minimal waste.
Say the movement is fast. We can use the case, locus towards 0 instead of upping FPS. The direct opposite. To implement the silicon, we need a dedicated fv co-processor or GPU with circa 4 cores or more depending on variables.
To train AI, it is the same but with a co-processor which matches fv to other variables, the vector being a linguistic, photo, scientific (physics), or logic problem. To achieve processing, the subject is reverse taken apart (reverse engineered) by fv, then compared to another, fv‘ and so on. There is no need for specialized software and hardware.
...
Finally managed to unload / get cleaned. 1 1/2 hours till chores. My throat is less sore and am feeling less sleepy.
Am wondering if the above 'modern computer architecture' is any good. I wrote it last night as I was going under my sedative. It might not be very correct, but I think the gist of it is useful and someday computer GPUs may be cheaper / more efficient this way.
As for the CPU itself, there are just 3 basic functions it needs to handle which are fr , fw , and fop. The logic behind this is to simplify and modularize reads, writes, and operations. Since we deal with functions of r, w, and op, we do not need a complex CPU but 3 simple cores that carry their own instruction local sets, hidden from one another.
For example fr may be tasked by the program, to read AddressN, write to AddressN+1, remove AddressN through an operation of N NOR N+1. It need only take 'read AddressN' and pass the code to the next core. The function of r, will handle the execution async and in OOP mode. Therefore fr needs to have it's own scheduler-listener-passer as do the other cores.
Such a CPU is easy to build and relatively fast for the cost. It is easy to code for and can scale perhaps to 100s of cores on a single chip.
