What Will Make Computers Faster: An Approach Based on Computational Complexity
Владик Крейнович · scholarworks - UTEP (The University of Texas at El Paso) · 2016
Why do we want computer to become even faster?Computers are very fast already.Yes, they are getting faster year after year, but, honestly, other than making videos more realistic and faster to download, from the purely scientific viewpoint, there is no big reason to replace a 5-year old computer with the new one: already on a 5-year old computer, we can do all the editing and all simple computations very fast.However, there are practical problems in which it is desirable to make computers much faster.For example, we now have reasonable algorithms for predicting weather, algorithms that are much better in predicting tomorrow's weather -and even weather for a weak ahead -than a random guess.There are even algorithms that can reasonably well predict the path of a tornado.However, while algorithms for predicting tomorrow's weather take less than a day on modern high performance computers, algorithms for predicting where the tornado will be in 15 minutes requires several days of intensive computations.This large computation time defeats the purpose: we will learn where the tornado moved before the computations are over.How can we make computers faster: traditional engineering approach.Traditional approach to making computers faster is based on introducing newer and newer engineering ideas:• sometimes, a new material can help, • sometimes, a new technology for manufacturing these computers, etc.Limitations of the traditional approach.The traditional engineering approach to speeding up computers have worked reasonably well since 1960s: the resulting com-