Thursday, November 03, 2011

The Technology of Calculation

We've come a long way since the days of wearing animal skins and counting on our fingers.

We graduated to notches on sticks, knots on a string, and piles of stones, then moved up to the next great technological advance - the abacus.

In time, we moved on to the slide rule, adding machines, and pocket calculators ... my first "pocket calculator" was too big to fit into any pocket not owned by Captain Kangaroo. It added, subtracted, multiplied, divided, and calculated reciprocals. It had no memory. It weighed about half a pound and cost $98. Today, the cheapest of pocket calculators has more computing power than mission control for the Apollo moon landing, more memory than your vindictive ex-wife, and is small enough to get lost in the lint at the bottom of an average pocket.

Then came computers.

Then came supercomputers.

And then came the need to devise a whole new vocabulary to describe how fast those supercomputers are: not just the number of calculations they can perform per hour or minute, but the number of floating point operations per second (or flops) they can perform ... and not just flops, but teraflops (one trillion flops) and petaflops (a thousand trillion flops).

Those babies are fast.

Which leads me to this article from yesterday's Washington Post: Supercomputers Offer Tools for Nuclear Testing — and Solving Nuclear Mysteries. You may recall my post from this past Monday that discussed the dismantling of the last B53 thermonuclear bomb in the American arsenal ... the bomb that had the explosive power of 9 million tons of TNT, would leave a crater 750 feet deep, and kill everything within ten miles.

As you might imagine, such a bomb is relatively difficult to test. The neighbors tend to complain when you detonate 9 million tons of TNT in your back yard. And then there was that pesky Comprehensive Nuclear Test Ban Treaty (which the US has never ratified, but with which it nevertheless complies). So, enter the supercomputers, which can test the viability and power of enormous bombs mathematically. This is a picture of the supercomputer named Dawn (IBM BlueGene P) at the Lawrence Livermore National Laboratory:

Well, actually the jury is still out on whether or not supercomputers can accurately test nuclear weapons, according to the Post article. It seems the only way you can really know if the bomb works is to drop it on something and see what happens. Which isn't always a good test plan.

You'd think that we'd have better things to do with all that computing power than figure out better ways to kill each other. Perhaps a supercomputer could replace all those economists that can't agree on why nobody has jobs. Or do the work we thought we'd elected Congress to do. Or just play chess really fast and really well. Anything but figure out whether that crater left by a random B53 will be 750 feet deep, or just a measly 500.

As for me, I'll just stick with my trusty pocket calculator. As long as it helps me know how much is in my checking account to the nearest $5 or so, I'm good.

And I don't think we ought to waste time building and testing nuclear bombs, anyhow. I like my world to be not dotted with 750-foot-deep craters.

Have a good day. Calculate accurately. More thoughts tomorrow.



eViL pOp TaRt said...

In effect, this is not much different from 1945 except for the magnitude. I read that when they first tried to test the atomic bomb, they did not now how much of an effect it would have. Scary!

Mike said...

'a supercomputer could replace all those economists'

A super computer can't actually think...... which means it's probably smarter than all economists put together.

Wv: blesoli - Oli must have sneezed.

allenwoodhaven said...

Don't go through life being crater-phobic. You can cure it by starting small, say with a shovel, and working your way up.