(2023-05-10) The perfection of imperfection
-------------------------------------------
One of my upcoming pet projects, as crazy as it might look, will have a
profiling part to determine the approximate amount of cycles that we need to 
skip on a particular machine to fit into a single 1/60 second frame. And I 
want to stress on the word "approximate" here. First, the timing method is 
itself imperfect (the sacrifice needed for full POSIX compliance) and only 
returns two figures after the decimal point. Second, the calculated result 
may vary between different runs and really depends on the current system 
workload. Third, we don't need the exact cycle count there, only the order 
of magnitude by which we need to slow down every frame loop iteration, and 
on the slower systems/runtimes, the amount of cycles we need to skip may be 
10 to 30 times less than on higher performance systems. Regardless of the 
project itself, this alone is a very interesting challenge. What's more 
interesting is that the exact precision is not possible to achieve and not 
generally required.

As another example, I have a circular slide rule (model KL-1) in my pocket
(holstered in a small pouch that originally belonged to a cheap Chinese 
pocket microscope). Two figures after the decimal point is the precision 
limit it can operate with, and even that can't be achieved in all possible 
conditions. Does this make it useless? Hell no! It's an efficient tool to 
quickly estimate square roots, sines/arcsines, proportions, reciprocals and 
products. And if I can't measure length with resolution less than 0.1 mm, 
weight less than 0.1 g and time less than 0.5 s anyway, why would I need to 
calculate anything using these measurements with the precision less than 
0.01? If I, for instance, use my scientific calculator and get the result of 
7.8378332, it will be of the same value to me as the 7.84 I'm gonna get on 
the slide rule. Because the source measurements are just as imprecise to 
make the rest of decimal digits insignificant.

What I'm trying to say is that there is a whole bunch of engineering tasks
that don't require us to operate with exact numbers. Just because we got 
used to them in the today's world doesn't mean we need them every single 
time. And this has its own beauty in it, as well as the fact that more older 
and reliable tools are still relevant to this day.

--- Luxferre ---