Your right, there is no one single theory for parallel computing... There are SEVERAL. There is a whole field dedicated to the subject that is several decades old. It is crazy to suggest that just because one we don't have one single equation that describes a whole field of study, that we don't understand that field. That is what you arguing, and buy that standard we know nothing about physics, math, and really every science out there. In other words, you don't believe there is such a thing as science because we don't have a single math equation that explains a whole field of science.
Bullshit. I've only argued that it isn't a mature science, not that it is merely so much voodoo. When you have 20 different theories and 400 different ways of approaching the same subject its absurd to claim it is a mature science. Right now heterogeneous computing is right up there with psychology in that respect. Both can definitely be called sciences, but only a fool would call them mature sciences just because they've been around awhile and continue to produce useful results.
Again, what does this field need to prove to you that it is proven? Computer science is one area that has WAY more math backing it than any other branch of science. We have probability theories that describe what will happen and we have exact theories that will describe exactly what is going on.
We can also use psychology to describe countless things too and make accurate predictions, but that does not make it a mature science. If you can't understand the difference, then there is no point in my elaborating further.
The very act of writing a computer program is to describe exactly what is going to happen in a given situation. You act as if we just throw stuff at the computer and hope everything turns out well. We don't.
Computer science is based off of math and as an extension parallel programming. In is the most solidly math based study out there. To make the assertion that we have no math theorems to back it is crazy. The entire field is a collection of math theorem after math theorem to describe exactly what is going on. Computer Science is the only science out there that depends ENTIRELY on math in order to work.
Do you consider Physics, Computer science in general, math, geometry, Biology, Chemistry or Horticulture to be a known science? Give me a field you consider to have a stronger scientific and math backing than parallel computing.
You're confusing what are called the "exact" sciences with "mature" sciences. Wine tasting is a mature science, but not an exact science. Chaos theory is an exact science, but not a mature science. One is not synonymous with the other.
CPU processor architecture I consider to be a pretty mature science at this point. They have a very solid theoretical foundation and we've even made significant strides in multicore processor architecture. Heterogeneous architecture is another matter altogether.
Sure, we've had some experience with parallel processing and heterogeneous architecture, but by some estimates single chip heterogeneous computing will eventually require 4 distinct types of cores to maximize their speed and power. The problem is we just don't know what the best arrangement might be and have no real theory to turn to for answers. For that matter, it might be possible to automate multicore threading in hardware right on the chip, but NOBODY KNOWS.
These are both basic questions that we simply don't have the theory or experience to answer because the science is immature. Not because the science is not mathematical or rigorous or empirical or cutting edge. Simply because it is immature.
Facebook
Twitter