Test time!!

How fast is fast? Please answer by explaining an Exaflop in your OWN words. Explain what that means for computing.

The Flops component of the name stands for FLoating point OPerations per Second. The full sequence and accepted numerical value in computing terms is as follows.

Kilo (2^10) - 1,024
Mega (2^20) - 1,048,576
Giga (2^30) - 1,073,741,824
Tera (2^40) - 109,951,162,800
Peta (2^50) - 112,589,990,700,000
Exa (2^60) - 115,292,150,400,000,000

So let's say your programing on your brand new Exaflop computer. What types of programs would you execute?

Begin Universal Simulation...
Simulation ready...
Initiating Neural transport stack...
Ready...

Which entity would you like to enter Universe as?

God User
Random Human
Other: Please specify

And for those not knowing what Floating point is?

Scalars are numbers like 1,2,3,4,5 ... and are easily stored as binary as 001, 010, 011, 100, 101 and so on. So to add two numbers 1 + 1 say you add 001 and 001 = 010. But what about 2x2 = 4. This is 010 x 010 = 100 or just add two to itself twice. The same for any basic operation.

But how do you store 1.1 or do 1.2 x 3.5? That is a floating point operation. One trick is to say 1.1 = 01.01 or, if you have 8 bits to play with you use the first 4 as mantissa and the second four as the decimal point so 1.1 is 00010001. But that limits you to a maximum of 8.8. The trick is to use several binary numbers to represent the digits. But that requires a lot more processing to handle. Important in graphics and scientific calculations.

You need more registers to store numbers and better designed hardware. It used to be that x86 processors where scalar only and you paid extra for the floating point processor. Hence the old 486SX/DX variants.

And as the man says Exa is 2^60

Give me Linux OS/390 running on 4 tightly coupled IBM mainframes and I'll give you fast. Or 20 Alphas with 64xEV7 CPU's, each.